Final Regulatory Impact Analysis and
Summary and Analysis of Comments on the NPRM
Interim Control of Gasoline Volatility
January 19, 1989
U.S. Environmental Protection Agency
Office of Air and Radiation
Office of Mobile Sources
-------
Table of Contents
Page
Introduction
1. Period and Location of RVP Control 1-1
2. Effect of RVP on Vehicle Emission Factors 2-1
3. Environmental Impact 3-1
4. Economic Impact on Refineries 4-1
5. Cost Effectiveness 5-1
Appendix
Summary and Analysis of Enforcement Comments
-------
This Final Regulatory Impact Analysis (FRIA) comprises a
significant portion of the supporting documentation for an
interim national RVP control program, reducing gasoline
volatility to 10.5, 9.5, and 9.0 psi, depending on the area of
the country. Documentation not included in this Final RIA may
be found in the Draft RIA associated with the August 19, 1987
volatility proposal or in the Preamble to the Final Rule for
the interim RVP control program.
Included in the analyses found here are summaries of
comments received on the issues pertaining to the choice of the
period and location of RVP control, the cost, the environmental
impact and the cost effectiveness of an interim control program
cind EPA's responses to them. Also included is a summary and
analysis of comments received on how to treat alcohol blends
under such a program. An -appendix to this document contains
the summary and analysis of comments on the enforcement aspects
of this program. The Preamble addresses comments on the impact
of gasoline volatility regulations on the natural gas liquids
industry, the alcohol blend industry and small refiners.
Each chapter of this Final RIA begins with a brief
synopsis of the methodology and significant input factors of
the Draft RIA analysis. This is followed by a summary and
s.nalysis of the comments received. Completing each chapter is
the final analysis of that aspect of the regulatory analysis.
-------
Chapter 1
Volatility Control Period and Gasoline Distribution
I. Synopsis of NPRM Analysis
EPA's volatility proposal which would limit gasoline
volatility during a determined regulatory control period was
based on various analyses and assumptions. The length of the
control period was decided upon after examining the seasonal
pattern of ozone violations and the transition times necessary
to blend down the high RVP gasoline to compliance RVP levels.
Volatility control levels were based on the ASTM RVP class in
each state. The relative RVP control levels for each ASTM RVP
class were based on emission analyses showing roughly equal
emissions when constant percentage reductions were applied to
the ASTM limits. The following paragraphs summarize the
analyses performed and the conclusions made.
A. Control Period
Ozone violations in non-attainment areas follow a seasonal
trend, with the majority of violations occurring during the
summer months. This is due to the higher temperatures and
longer hours of daylight, both of which contribute to ozone
formation. Because of this seasonal trend, hydrocarbon (HC)
emission control measures which have the flexibility of being
implemented for only a few months during the year are
advantageous. Volatility control is one such HC emission
control measure which can be implemented for only a few months
during the year as needed. In order to determine the best
period to implement this control, EPA in the DRIA examined both
the trends of ozone violations in non-attainment areas and the
time required by the gasoline distribution system to deliver
RVP-controlled fuel to service stations.
1. Seasonal Pattern of Ozone Violations
The analysis which EPA used in the proposal to determine
the period during which nationwide ozone violations occurred
focused on the "design value day" of the non-attainment areas.
This day is defined to be the day on which occurred the fourth
highest daily maximum one-hour measured ozone concentration in
the area over a three-year period (or, when less data is
available, the third highest from a two-year period or the
second highest from a single year's data). The analysis,
therefore, focuses on a one-day period for each non-attainment
area. EPA's analysis in the DRIA of seasonal ozone
non-attainment examined the distribution of design value days
over the twelve months in order to determine a period during
which nearly all of these days would occur. The analysis used
1982-84 design value days for non-California urban
non-attainment areas.
-------
1-2
EPA also performed a second, broader analysis in the DRIA
which examined all ozone exceedance days rather than just
design value days. This analysis focused on 1981-83
exceedances rather than 1982-84. The total annual exceedances
by month for the non-attainment areas were tabulated. Because
exceedances were taken from each monitoring site an;;, a
non-attainment area may have more than one monitoring site, any
specific non-attainment area could have more than 31
exceedances per month (one violation per day) depending on how
many monitoring sites the area had.
Finally, EPA evaluated seasonal trends in peak ozone
concentrations. The average of all areas' maximum ozone levels
by month were used to determine over what period peak
concentrations occurred.
Houston was singled out as a having a unique, year-round
ozone problem and some of the analyses excluded Houston due to
this fact. The ozone exceedance data for Houston showed a
relatively large number of exceedances in 1980 and 1983. Also,
Houston recorded exceedances in every month of both years; no
other area recorded exceedances in more than eight months of
either year.
Based on the results of the above evaluations, EPA found
the two primary choices for seasonal ozone control to be
May-September and June-September. Because of the greater
potential to impact ozone exceedances, May-September was
recommended as the period when ozone control should" be
regulated.
2. Transition Times for Compliance
The proposed volatility control program was aimed at the
months during the year when the majority of ozone exceedances
occur: May through September. EPA recognized the fact that in
order to be in compliance at all points in the distribution
system, fuel must be refined and shipped before the compliance
date. Based on a contractor report ("Petroleum Storage and
Transport Times," Jack Faucett Associates (hereafter Faucett
Study)) and EPA evaluation, the Agency assumed that there would
be an average shipping time of about four weeks prior to the
beginning of the control period and four weeks at the end.
3. Starting and Ending Dates for the Regulatory Control
Period
May 16 through September 15 was proposed as the regulatory
control period and was intended to provide control during the
entire May through September period. This took into account
the calculated transition times at the beginning and end of the
control period. It also assumed that for the estimated four
week transitions at the beginning and end of the control
-------
1-3
period, controlled fuel would increasingly enter the market,
resulting in the equivalent of • two weeks of complete control
during each four week transition period.
B. State-by-State Volatility Classifications
In the Volatility DRIA, EPA proposed to adopt RVP control
levels which represented a constant percent reduction from the
current ASTM gasoline volatility classification system for each
state or portion of a state during the May through September
period. This was based on the observation that such a system
yielded roughly consistent evaporative emissions in the 61
non-attainment areas. From the viewpoint of gasoline
distribution, this was also the most straightforward system to
use due to the fact that participants in the distribution
system were already familiar with the classification system.
C. Relative RVP Standards in the ASTM Class A, B and C
Areas
Proposed volatility standards in the NPRM were derived
using Class C fuel as a base. Reductions in RVP levels
proportional to the Class C reduction were then applied to
Class A and B areas in order to obtain equivalent degrees of
omission control in all areas of the country. Class C areas
were used as the basis since the climate implicit in the
Federal Test Procedure (FTP), used in developing our test data
on the effect of RVP on emissions, most closely matches the
climate in Class C areas.
II. Summary and Analysis of Comments
EPA received comments on each of the topics summarized
above. The following sections will summarize the comments
received in these areas and give EPA's response to these
comments.
A. Control Period
Comments on the control period were divided into three
main topics: 1) the seasonal pattern of ozone violations; 2)
the length of time required to make the transition to the lower
RVP fuel and problems that may occur during this transition;
and 3) the starting and ending dates of the control period
itself.
!• Seasonal Pattern of Ozone Violations
Although EPA proposed a four-month period for volatility
controls (mid-May to mid-September) which when combined with
the transition periods was intended to give sufficient control
for the five month period of May through September, NRDC
pointed out that ozone violations are not solely confined to
-------
1-4
the summer months. They commented that EPA data show that
nineteen cities experienced at least one ozone violation
outside the May-September period in 1980. Because of this,
there should also be concern about preventing ozone violations
in early spring and "Indian summer."
SOHIO stated that if the necessary two-month transition
periods were added to the proposed regulatory control period,
the length of the control period would increase. This,
however, would have a minimal impact on. the ambient ozone
levels since the emission reductions would occur before or
after the "ozone season."
Whereas the DRIA analyses suggested that ozone control
during May through September was most important, EPA has
reanalyzed this issue using more recent ozone data. Ozone
exceedance data from 1985-87 was examined to determine more
precisely when ozone exceedances occurred. Analyses were
performed to determine the percentage of exceedances that
occurred during the proposed period of May through September,
and then to look at the percentage and level of exceedances in
April and October. The geographic locations of April and
October exceedances were then examined to determine whether the
pattern of exceedances was unduly influenced by any unusual
local ozone problems.
Table 1-1 shows the results of an analysis which
calculated the percentage of ozone exceedances which occurred
during various monthly periods. Due to the fact that
California and Houston both have serious ozone exceedance
problems during many months of the year, the analysis was
performed both with and without these areas. Their unique
ozone problems influence the potential control period and will
likely warrant extra local control strategies in both cases.
EPA found that for the entire nation 88.5 percent of the
exceedances fell in the May through September period. This
increased to 95 percent when California and Houston were
excluded.
EPA then looked at the months of April and October to
determine how frequent were the exceedances in those months.
For the entire nation, 4 percent of all exceedances fell in
each of the months of April and October. When California and
Houston were excluded, 3 percent of the exceedances occurred in
April and only one percent in October. These results show that
control during the period of May through September would affect
approximately 95 percent of the nationwide exceedances.
In addition to looking at the frequency of the
exceedances, the ozone levels associated with the 1985-87
exceedances in May and September as well as April and October
were examined. At an ozone concentration greater than or equal
-------
1-5
Table l-l
1985-87 Ozone Exceedances by Month (Percent of Total)
Entire Nation Without CA and Houston
April 4 3
May-September 88.5 95
October 4 1
Percent of Exceedances Greater than or Equal to .14 ppm 03
April 45 36
May 38 32
September 44 22
October 69 67
-------
1-6
to 0.14, additional control over the stationary and mobile
source control which is currently being phased in may be
necessary to bring an area into compliance. Table 1-1 shows
the percentages of exceedances greater than or equal to 0.14
for each of the four months. In April, 45 percent of the
exceedances for the entire nation were greater than or equal to
0.14. When California and Houston were excluded, this dropped
to 36 percent. For comparison, in May 38 percent of
exceedances nationwide were greater than or equal to 0.14.
Excluding California and Houston, the percentage became 32
percent. In October for the entire nation, 69 percent were
equal to or above 0.14. This only dropped to 67 percent when
California and Houston were excluded. In September 44 percent
of nationwide exceedances and 22 percent without California and
Houston were greater than or equal to 0.14.
The uncontrolled diurnal index (UDI) is another measure of
the desirability of applying RVP control to reduce the level of
ozone exceedances, in this case evaluating the likely magnitude
of evaporative emissions in an area. The indices are
calculated using the actual temperature range of the city on
the exceedance date, the atmospheric pressure of the area based
on its elevation, the RVP level of the area based on its ASTM
classification, and an assumed 61.5 percent full gas tank to
account for fuel weathering and in-use tank level variability
(see Chapter 2 of DRIA).
UDIs were calculated for the 1985-87 exceedances in April
and October. A UDI equal to 2.0 represents the UDI level for
11.5 RVP fuel under FTP certification temperature and tank
level conditions (representative of current ASTM Class C
areas.) Areas with a UDI greater than 2.0, therefore, would
likely be in need of control. UDIs were calculated for the
April and October exceedances in order to determine the degree
that evaporative emissions were contributing to these ozone
exceedances. For the April exceedances, 58 percent were
greater than 2.0, with an average UDI of 2.793. Similar
results occurred in October, with 62% of UDIs being greater
than 2.0, and with an average UDI of 3.311.
Both the UDI and ozone level analyses indicate that ozone
control during April and October requires serious
consideration. However, it must be remembered that only 3 and
1 percent of all ozone violations occur in April and October,
respectively. Thus, while the violations are often severe and
excess evaporative emissions are high, the violations are
infrequent. Also, the dates on which these April and October
exceedances occurred show that control during only a portion of
each month may be sufficient. The majority of exceedances in
April occurred during the last two weeks (79 percent), and in
October 84 percent occurred during the first two weeks.
Therefore, beginning the control period May l would likely
-------
1-7
provide substantial control during the last two weeks of
April. While ending control September 15 may only provide
slight control in early October, the small fraction of ozone
violations occurring during October made the extension of RVP
control through September less efficient.
2. Transition Times for Compliance
Nearly all the comments received on this issue were from
refiners who stated that two months are needed once refiners
begin shipping low-RVP fuel to mix down the high RVP gasoline
•and be in compliance with the low RVP standard. API commented
that although only four to five weeks are currently needed to
make the transition, lowering the RVP even further would
dramatically increase the transition time necessary for
compliance. API estimated that 60-65 days will be needed for
95 percent of the low RVP gasoline to reach the consumer with
the more stringent regulations. API stated that the Faucett
study (which was used by EPA to determine the transportation
times involved in the transition) did not accurately describe
the transportation and storage times which would be experienced
in a more stringent volatility controlled environment such as
the one proposed by EPA.
After conducting their own survey of industry
transportation and storage times, API commented that the
estimated times of supplying gasoline presented by Faucett for
the current regulations appeared somewhat low but were not
unreasonable. However, the estimated transportation and
storage times under the volatility proposal were underestimated
by Faucett because a "plug flow" assumption was used to
describe the movement of gasoline through the distribution
system.
Amoco, in agreement with API, stated that the Faucett
study did not recognize back-mixing effects in storage tanks.
They asserted that Faucett's plug-flow assumption was overly
optimistic and that the transition times were, therefore,
underestimated. Amoco also stated that the transport times for
marine movements used in the study were low. Valero Refining,
in quantifying the effect of the proposed RVP regulations on
their operations, modified the Bonner & Moore linear
programming system to include a one month transition time as
input unlike the two months that others were claiming.
However, they also mentioned API's two month transition time as
the amount of time that may actually be necessary.
In addition to increased costs, the problems brought up in
the comments about the two month transition were problems of
safety, driveability and increased HC emissions. According to
comments, these problems would occur due to low RVP fuel
reaching the consumer early in the transition period when there
may still "be cooler temperatures.
-------
1-8
In response to these comments on EPA's transition
assumptions, the transition times were reanalyzed in greater
detail in order to take into account backmixing effects. This
reanalysis used the transportation and storage times from the
Faucett study as a basis in the analysis. Although some
commenters felt these numbers weren't accurate, the Faucett
estimates were in general agreement with API's values. In any
event, the commenters did not provide sufficient justification
for EPA to revise the Faucett values.
The goal of the reanalysis was to calculate the 70th
percentile transition time for a terminal to mix down to
compliance from an unregulated Class C RVP to a regulated Class
B RVP. The 70th percentile time was chosen because refiners
have some control over the distribution system, being able to
send one RVP fuel to some regions while sending another
elsewhere. However, this control is not absolute and some fuel
must reach certain sites early for another to just be in
compliance in time. The calculated transition time was an
average of transition times for a one-terminal and a
two-terminal line (in a two-terminal line the gasoline goes
through two terminals before going to service stations). The
analysis assumes 70/30 mixing (70 percent new fuel, 30 percent
old fuel) for the one-terminal line, and 90/10 and 70/30 mixing
for the two terminals in a two-terminal line, respectively.
Terminals were assumed to receive new batches of fuel every
five days for the first terminal, and every seven days for the
second terminal in the line. (One-terminal lines are assumed
to receive fuel every five days.)
The transition time includes the transportation time from
the refinery to the first terminal by pipeline (9.4 days
according to Faucett), the average amount of time necessary for
the first terminal to mix down to compliance after the first
batch of new fuel is received (15 days for a one-terminal line,
5 days for a two-terminal line); the average additional amount
of time necessary for the second terminal to mix down to
compliance after the first terminal is in compliance, if
applicable (22 days); the extra amount of time needed for the
70th percentile terminal to come into compliance to account for
the fact that all the base values from Faucett were for the
50th percentile terminal (5 days according to Faucett); and,
finally, seven additional days to account for any delays.
Thus, an average of about six weeks appears to be needed for
terminal compliance at the beginning of an RVP control period.
For service stations, it was assumed that there was 80/20
mixing in the tanks and, based on the Faucett study, new fuel
was received every three days. From this, an average of an
extra five or six days will be needed, above the terminal
compliance time, for service stations to come into compliance.
-------
1-9
Transition times were also calculated for the situation of
an additional mid-summer RVP change (i.e., from a Class C to a
Class B or a Class B to a Class A requirement) . The same
assumptions as in the above calculations were used in this
analysis. In this case since the RVP of the fuel has a shorter
range by which to decrease, the transition time would be
shorter than that for the first change of the control season
and would take about 5 weeks for terminal compliance.
At the end of the control period, EPA determined in a
similar analysis that the refineries can on average stop
producing the controlled fuel 8 days before the end of the
control period. This is the amount of time necessary, based on
the Faucett study, for the 30th percentile terminal to receive
its first batch of uncontrolled fuel from the refiner. This is
not the same amount of time as at the start of the control
period because, after this first batch has been mixed in at the
terminal, the terminal is no longer in compliance with the
lower RVP standard.
Therefore, when accounting for backmixing in the tanks,
EPA concluded that under the proposed volatility regulations
the required transition time at the beginning of the control
period would be six weeks rather than the two months the
comments suggested. In addition, refiners can stop production
an average of one week early.
Finally, the potential problems raised by the commenters
on safety, driveability, and increased HC emissions during the
transition period only apply to the second proposed phase of
RVP control and are not a factor for the first phase of control.
3. Starting and Ending Dates for the Regulatory Control
Period
EPA proposed a control period of May 16 through September
15. In response to this, various other periods were suggested
in the comments. Amoco suggested a two-month regulatory period
of July 1-August 31. They reasoned that when adding the sixty
day period needed to flush the distribution system, four months
of control would result. NRDC and Chrysler both stated that
the period of control may need to be extended to a five month
period of May 1 - September 30. NRDC commented that because
ozone violations also occur outside the summer months,
year-round controls on volatility may be desirable to the
extent that they are consistent with driveability
considerations. NRDC suggested that at a minimum, the control
period should be extended into the warmer periods of the spring
and fall. Chrysler suggested that this longer control period
is preferable because many southern areas have summer-like
temperatures before May 16 and after September 15. They also
commented that this period would be less disruptive to current
-------
1-10
procedures of gasoline distribution. The New Jersey Department
of Environmental Protection commented that the proposed season
of May 16 through September 15 is too narrow to satisfy the
needs of the Northeastern United States. They stated that the
Northeast often experiences exceedances during April, early
May, and late September. They suggested that an intermediate
limit for volatility could be set for these time periods. Sun
Oil Co. and NADA both stated that the control period had been
appropriately limited. NADA commented that since very few
ozone violations occur during the winter months, EPA is
justified in focusing on the four month period of mid-May
through mid-September. Volkswagen suggested that a longer
control period (beyond May 16-September 15) should only be
evaluated after this regulation has been implemented to see if
an extension is warranted.
In order to respond to the comments received and to
determine the appropriate regulatory control period, the amount
of control obtained during the transition period and the RVP
levels at terminals and service stations during the time period
of March through mid-November were reanalyzed.
Using the calculated transition times, RVP levels at
terminals and stations over the summertime period were
determined for different regulatory control periods. Two cases
were examined: one for a terminal compliance period of May 1
through September 30, and a second, for a terminal compliance
period of May -16 through September 15. In order to determine
the RVP level obtained during the transition periods at the
beginning and end of the control period, several assumptions
were made. The partial control RVP levels were calculated
assuming 70-30 mixing at the terminals with seven days between
new batches of fuel and one-seventh of the total number of
terminals receiving the new batch each day. For service
stations, 80-20 mixing was assumed, with new batches coming
every three days and one-third of the stations receiving the
new fuel each day.
As shown in Tables 1-2 and 1-3, this analysis showed that
a terminal compliance period of May 1 to September 30 gives, on
average, full compliance at service stations from May 8 to
September 30 and partial control beginning around April 1 and
continuing as late as November 8. A terminal compliance period
of May 16 to September 15 gives full station compliance from
May 23 to September 15 and partial control beginning around
April 15 and continuing as late as October 22. While the
tables show the results of a reductions to 9.0 RVP, the times
to blend down to 10.5 RVP are the same.
The results of this analysis show that when mixing down
from an uncontrolled to a controlled RVP level, 85 percent of
the RVP reduction at the service stations has been achieved two
weeks prior to the start of the terminal control period. In
mixing from a controlled to an uncontrolled level at the end of
-------
1-11
Table 1-2
Transition RVP Levels
Date
March
April
1
15
30
1
8
15
22
30
1
8
15
30
May
June
July
August
September 1
15
30
October
November
1
8
15
22
30
1
8
15
Compliance
Refinery
13.50
9.00
9.00
9.00
9.00
9.00 '
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
13.50
13.50
13.50
13.50
13.50
13.50
13.50
13.50
13.50
Period: May 1 -
Terminal
13.50
13.50
13.50
12.86
10.22
9.36
9.11
9.03
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.45
12.28
13.14
13.39
13.47
13.50
13.50
13.50
13.50
Service Station
13.50
13.50
13.50
13.33
11.10
9.65
9.21
9.05
9.04
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.12
11.38
12.85
13.29
13.45
13.46
13.50
13.50
Assumes transition from uncontrolled Class D fuel (13.5)
to controlled Class C fuel (9.0).
-------
1-12
Table 1-3
Transition RVP Levels
Date
March
April
May
1
15
30
1
15
16
23
30
7
14
15
23
30
June
July
August
September 1
15
16
23
30
October
6
14
15
22
30
ial Compliance:
Refinery
13.50
13.50
13.50
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
13.50
13.50
13.50
13.50
13.50
13.50
13.50
13.50
13.50
13.50
May 16 - S«
Terminal
13.50
13.50
13.50
13.50
. 13.50
12.86
10.22
9.36
9.11
9.03
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.45
12.28
13.14
13.39
13.47
13.50
13.50
13.50
Service Stat^
13.50
13.50
13.50
13.50
13.50
13.33
11. 10
9.65
9.21
9.05
9.04
9.00
9.00
9.00
9.00
9.00
9.00
9.00
9.12
11.38
12.85
13.29
13.45
13.46
13.50
13.50
Assumes transition from uncontrolled Class D
to controlled Class C fuel (9.0).
fuel (13.5)
-------
1-13
the control period, 85 percent.of the control at the service
stations is lost two weeks after the end of the control
period. Therefore, a control period beginning May 1 at
terminals would provide two weeks of additional, nearly
complete control at service stations at the end of April and
some control in early April. A control period ending on
September 15 at the terminals, however, would provide little
added control at service stations in October.
In the DRIA, April was shown to contain more design value
days (4) than either May (0) or October (1). Also, April
contained at least the same percentage of ozone violations (1
percent, excluding Houston and California) as October (0-1
percent); May contained more (1-3 percent). The more recent
analysis shown in Table 1-1 here indicates the same: April
contains a greater percentage of ozone violations (3 percent)
than October (1 percent). Though a greater percentage of
October's violations are over 0.14 ppm ozone, its total number
of such violations is lower, due to its lower total number of
violations. Therefore, it appears more important to obtain
some control in late April than in early October and a
regulatory control period of May 1 through September 15 for
terminals is recommended.
2. State-by-State Volatility Classifications
As stated above, EPA proposed to enforce the current ASTM
monthly volatility classifications and geographical boundaries,
but with proportionally reduced RVPs during the control
period. The one change to the ASTM system that EPA proposed
was to eliminate the transition months, requiring compliance
with the lower volatility class during months with two
classifications. Marathon supported compliance in accordance
with the ASTM schedule of seasonal and geographical volatility
classes, stating that a different EPA-designated control period
should not be required. Phillips supported this, as well, but
stated that the current ASTM transition periods should be kept
since eliminating them would significantly increase costs at
the beginning and end of the ozone season.
GARB did not agree with enforcing the ASTM monthly
classifications or geographical boundaries as they apply to
California. They commented that, for California, the EPA
system should match California's current requirements.
California is currently divided into fourteen air basin areas,
with five different ozone seasons. The seasons are all
contained in the period of April-October. Volatility is
limited to 9 RVP during the given ozone season. GARB commented
that EPA's interim proposal for California is not as stringent
as their current system and that any regulation should be at
least as stringent as the current one.
-------
1-14
Other commenters held various opinions as to where
RVP-controlled fuel should be required. Sinclair commented
that the control should only be required in areas where it is
needed (i.e., ozone non-attainment areas), and, therefore,
should not be imposed uniformly on all refiners. Sinclair
suggested that either all gasoline refined in non-attainment
areas be required to meet "ultra-low" RVP specifications or
that RVP exemptions should be granted for gasoline shipped to
attainment areas. Sinclair also suggested that since the
Rockies and Plains were close to being in full compliance with
the standard, they may not need the proposed regulation. Since
only 40 percent of the fuel in the area is brought in from
outside areas (namely, Texas and Louisiana), the regulated
outside fuel would give enough of an RVP reduction to put the
area in full compliance. SOHIO stated that RVP control must be
enforced in all areas since refiners cannot segregate the
gasoline between attainment and non-attainment areas.
According to SOHIO, only segregation over broad regional areas
is possible. (For this reason, SOHIO suggested that EPA should
exercise all authority available to preempt and prevent state
volatility regulations.)
API, in agreement with SOHIO's comment above, stated that
the supply system is unable to differentiate between states, or
parts of states, with different ASTM classes. As a
consequence, they say, the lowest RVP gasoline for a given area
must be supplied to the entire area. TOSCO also stated that
the development of a patchwork of different volatility controls
across the country would create uncertainty for refiners and
marketers working to meet a variety of standards. This could
result in gasoline supply problems in certain areas of the
country. Wickland Oil raised a concern about the fungibility
of gasoline. Since the gasoline distribution system is based
on fungibility, they said, it cannot simultaneously handle
large quantities of the same grade of gasoline with different
RVPs. Wickland Oil currently sends 9.0 RVP gasoline to all
areas which they supply, although California, Nevada, and
Oregon have different requirements. Amoco commented that there
may be driveability problems in those Class C areas adjacent to
Class A areas (Denver, El Paso, Phoenix, Salt Lake City).
According to Amoco, some exemptions may be necessary to amend
this.
Volatility designations on a state-by-state basis were
reevaluated to determine if any changes to the current ASTM
class system should occur. (EPA does not refer here to a
change of the ASTM RVP classification system, but to an EPA
system whose RVP control levels are a function of the ASTM
levels for each class.) Dual classification areas (i.e.,
states which according to the ASTM system are assigned either
of two RVP classes in a particular month to allow for
flexibility during transitions) were first examined to
determine which single classification should be applied to the
area. For states with non-attainment areas, the UDI on the
-------
1-15
design value day was the determining factor in the choice of
class. For those states without non-attainment areas, location
with respect to surrounding class levels was the main factor.
A UDI analysis was also performed for single-classification
areas to determine whether any changes to the current ASTM
class system might enhance the workability of the control
program without the loss of non-attainment area emission
benefits.
In the NPRM, EPA proposed that in dual classification
areas the lower of the two RVP classes be the basis of the
standard. This approach was reexamined on a state by state
-basis to determine how much control these areas need during the
months of dual classification to achieve the same emission
reduction as other areas. For non-attainment areas which had
dual classifications in any of the months from May to
September, UDIs were calculated for the temperature conditions
on exceedance days in that month with gasoline at the current
ASTM RVP levels. Where the UDIs in a particular state were
above 2.0 (2.0 being the UDI level for 11.5 RVP at FTP
certification temperatures and tank level conditions), the
lower RVP class was generally chosen for that state in that
month.
For those dual classification areas which had no
non-attainment areas, geographic proximity to states with
non-attainment areas was the main factor in the decision about
the level of the RVP standard. The standard was chosen based
on the RVP of surrounding states and on the ease with which the
distribution system could get fuel to that state. Agreeing
that the gasoline distribution system cannot easily distinguish
ASTM boundaries, as many comments mentioned, EPA grouped the
class areas in various regions of the country as much as
possible for each month.
Another UDI analysis was performed to determine if any
changes should occur in the relative RVP control levels of the
three classes which were based on the current ASTM class
system. As with the dual-classification analysis above, UDIs
were calculated for 1985-87 exceedance days using the current
ASTM RVP levels. By examining both the level of the UDIs and
maps of the current ASTM system and pipeline networks, EPA
constructed a revised system of RVP classes across the country
which retains most emission benefits while improving the
manageability of the program for the gasoline distribution
system.
Table 1-4 shows a list of states (or parts thereof) whose
control levels have been relaxed and the number of months it
was relaxed. The majority of the revisions affect Northwest
states which contain no non-attainment areas (ID, MN, ND, East
OR, SD, East WA, and WY) . These areas are now considered as
being Class C areas even though ASTM classifies them as Class
C/B or Class B. Relaxations from Class A to Class B of
southern Nevada and southern California were performed for
similar reasons, since they also contain no non-attainment
-------
1-16
Table 1-4
States With Relaxed RVP Control Designations
Relaxation from Class B to Class C:
Montana (5)*
Idaho (5)
Wyoming (5)
S. Dakota (5)
E. Washington
E. Oregon (4)
Colorado (1)
N. Nevada (1)
Utah (1)
(4)
Nebraska (5)
Kansas (3)
Oklahoma (1)
Arkansas (2)
Missouri (2)
Iowa (2)
N. Dakota (4)
N. Carolina (1)
S. Carolina (1)
Georgia (1)
Alabama (1)
Mississippi (1)
Louisiana (1)
Tennessee (1)
Relaxation from Class A to Class B:
S. Nevada (5)
SE California (5)
Arizona (2) W.
N. New Mexico (1) S.
Texas (2)
New Mexico (2)
Number in () shows number of months relaxed
-------
1-17
areas. Arizona was relaxed from Class A to Class B during
September due to the fact that it would be the only Class A
state in the nation during that month. The "rest of the
non-California relaxations were dual ASTM classifications which
are now being treated as the higher of the two classes rather
than the lower. These relaxations were based on an analysis of
the UDIs for the exceedance days of that month using the two
levels of RVP.[1]
After examining California's exceedances, EPA concluded
that those California areas which are currently Class C or B/C
are in need of a Class B level of control. [1] Also, the
southeast area which is currently Class A during the period of
May through September contains no non-attainment areas and was,
therefore, relaxed to Class B. For these reasons, and also in
response to CARB's comment that the RVP control boundaries in
California should match their current boundaries, California
was designated as Class B across the entire state for the May
through September period. Table 1-5 shows all of the
classifications which were reexamined and the final result,
where revisions were made to the proposed control levels.
3. Relative RVP Standards in the ASTM Class A, B, and C
Areas
In the NPRM, EPA proposed an RVP limit of 9.0 psi in Class
C areas with proportional RVP reductions in Classes A and B in
order to obtain equivalent emission reductions in all three
classes. The majority of comments disagreed with this approach
and stated that the RVP levels should be based on
cost-effectiveness rather than proportional reductions. This
will be considered in EPA's analysis of the final RVP level.
Those opposed to proportional volatility reductions stated that
reductions outside Class C areas are not cost-effective. This
is considered in Chapter 5. Others stated that RVP levels
below 9.0 psi in Class A and B areas will also cause
driveability and safety problems, thus compromising the
performance of the vehicle. This is an issue for the final
analysis of the second level of RVP control.
SOHIO stated that there is insufficient data to indicate
that emissions are reduced when volatility drops below 9.0 psi,
citing some data which indicate that total exhaust emissions
increase as the fuel volatility drops below 9.0 psi. Again,
all this is not relevant in this analysis of the interim
program.
NRDC questioned whether tying volatility levels to ASTM
classes was appropriate since ASTM limits are keyed to motor
vehicle performance, not air quality. They suggested that EPA
examine whether lower RVP limits would be appropriate in some
areas.
Phillips disagrees with basing standards on reductions
from Class C RVP levels, stating that the test procedure most
closely matches Class B areas or possibly a worst case Class
C. Phillips suggested that test fuels be matched to ASTM Class
-------
1-18
Table 1-5
Exceptions to the Proposed System Based Only on ASTM Class
State
Alabama
Arizona
Arkansas
California:
North Coast
South Coast
Southeast
Interior
Colorado
Connecticut
Delaware
Georgia
Idaho
Illinois:
North
South
Indiana
Iowa
Kansas
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada:
North
South
New Hampshire
New Jersey
New Mexico:
North
South
New York
North Carolina
North Dakota
Ohio
Oklahoma
(— ii
May
Final
ASTM EPA
B/A
C/B
B/A
C/B
C/B
D/C
D/C
C/B
D/C
D/C
C/B
D/C
D/C
D/C
D/C
D/C
C/B
C/B
C/B
B/A
D/C
D/C
B/A
B/A
D/C
D/C
D/C
C/B
B
B
B
B
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
B
C
C
B
B
C
C
C
C
idicates no change from ASTM system)1
June July August September
ASTM
C/B
C/B
A
B
C/B
B
C/B
B
B
A
C/B
Final
EPA
C
B
B
C
C
C
C
C
C
B
C
ASTM
C/B
A
B/A
C/B
B
C/B
B/C
C/B
C/B
B
B
A
C/B
B
Final
EPA
B
B
A
B
C
B •
C
B
B
C
C
B
___
B
C
Final
ASTM EPA ASTM
B/C
A
B/C
A B A
A/B A
C/D
B/C
B C B
B/C B
B
B/C
C/D
C/D
C/D
B/C
B/C
B C B/C
B C B
A B A
C/D
C/D
A/B A
A/B
C/D
B/C
B C B/C
.
Final
EPA
C
B
C
B
C
C
C
C
C
C
C
C
C
C
C
C
B
C
C
B
C
C
C
-------
1-19
Table 1-5 (confd)
Exceptions to the ASTM Class System
State
Oregon:
East
West
Pennsylvania D/C
Rhode Island
South Carolina
South Dakota
Tennessee
Texas:
East
West
Utah
Vermont
Washington:
East
West
West Virginia D/C
Wisconsin
Wyoming
(— i
May
Final
ASTM EPA
D/C
D/C
D/C
D/C
C/B
_ —
B/A
C/B
D/C
D/C
D/C
D/C
D/C
C/B
C
c
C
c
c
___
B
C
c
c
c
c
c
c
ndicates no change from ASTM system)
June July Auqust
Final
ASTM EPA
C/B C
B C
C/B B
C/B C
__ _
B C
ASTM
B
C/B
B
C/B
_ __
B/A
B
— —
B
Final
EPA
C
B
C
B
_ _ _
___
A
C
C
Final
ASTM EPA
B C
B C
— « _ — _ _
___
A/B A
B C
___ ___
___
B C
September
Final
ASTM EPA
B/C
C/D
C/D
B/C
B
B/C
_ __
A/B
___
C/D
B/C
-_—
___
___
B
C
— — —
C
C
c
c
c
— — -*
B
-._-.
C
c
___
___
___
c
Whe:a ASTM system indicated a transition month (e.g., A/B), EPA proposed that
that area be considered in the lower class.
-------
1-20
B volatility levels and that gasoline volatility be limited to
ASTM's current levels.
In the NPRM, EPA demonstrated that adopting ASTM RVP class
standards with proportional RVP reductions in each class would
provide roughly equivalent emissions across the country. EPA
has performed a new analysis of expected evaporative emissions
to examine more specifically whether the proposed standards
should be adjusted in order to achieve our goal.
Essentially, an analysis of emissions on ozone exceedance
days was performed by comparing UDIs in Class A and B areas to
those in Class C areas for both diurnals and running losses.
This analysis examined the UDIs on a monthly as well as an
overall basis. The proposed proportional-reduction RVPs were
used in these calculations (9.0, 7.8, and 7.0). The
temperatures used to calculate the UDI for the diurnal analysis
were the actual daily high and low temperatures of the
non-attainment area on the exceedance date during 1985-1987.
To simulate running losses, a second UDI was calculated using
the daily high as the initial temperature and a temperature
15°F above this as the ending temperature. The 15°F represents
the tank temperature increase resulting from a single LA-4
trip. The mean UDIs for each class on a monthly basis were
then calculated. For Class A, a single city, Phoenix, has a
large effect on the mean UDI; calculations were made both
including and excluding Phoenix. The results are shown in
Table 1-6.
The UDI levels for both diurnals and running losses were
as high or higher in Class A and B areas than the levels in
Class C areas. Therefore, when comparing the UDI levels in A
and B areas to the level in C areas, this analysis indicates
that no increase in the relative RVP standard levels from the
proposal for Class B and A areas can be made if the goal of
equivalent emissions in all areas is desired. If any change
were to be made, a decrease for Classes A and B would appear
necessary. This is primarily an issue for the final level of
RVP control. However, it does demonstrate that 9.5 RVP control
in Class B areas should be quite beneficial.
-------
1-21
Table 1-6
Non-Attainment Area UDIs on Exceedance Days
Diurnals
Running Losses
Class Month Mean UDI
A
B
C
June
July
August
Total
May
June
July
August
September
Total
Total
1.8137
1.2181
1.0631
1.2990
0.9269
1.2383
0.9954
1.1548
0.9405
1.0863
0.9004
w/o Phoenix Mean UDI
1.6057 3.1366
1.2181 1.8338
1.0631 1.5920
1.2005 2.0497
1.1906
1.7865
1.5512
1.7169
1.3304
1.6112
1.5352
w/o Phoenix
2.4567
1.8338
1.5920
1.8088
-------
Chapter 2
Effect of RVP on Vehicle Emission Factors
I. Synopsis of NPRM Analysis
The estimation of the effect of fuel RVP on vehicle
emissions is very complex, especially that for evaporative
emissions. In the Draft RIA, EPA utilized two distinct types
of models to estimate evaporative exhaust and refueling
emissions.
The most complex model applied to the estimation of
diurnal emissions. Here a chemical equilibrium model was
developed to predict the relative effects of fuel RVP,
temperature and vehicle fuel tank level on fuel tank
emissions. The estimated fuel tank emissions were then
correlated against the diurnal emission test results from EPA's
in-use emission factor program to provide in-use emission
estimates over a wide range of climatic conditions. The model
was also used to generate suitable average in-use fuel tank
levels to account for fuel weathering, RVP and temperature
variability and the wide range of fuel tank levels occurring
in-use. The assumptions behind this model were shown to hold
when applied against the test results of a smaller number of
vehicles which were tested over 27 different combinations of
fuel RVP and temperature.
The models used to estimate hot-soak and exhaust emissions
versus RVP were much simpler. They primarily involved
correlating measured emission test results under FTP conditions
versus fuel RVP. These correlations were used directly to
represent emissions in Class C areas. For Class A and B areas,
it was assumed that their emissions were the same as those of
Class C areas when their respective fuel RVPs matched the ratio
of ASTM RVP limits. For example, emissions in all areas were
equal when Class A, B, and C fuel RVPs were 9, 10 and 11.5,
respectively, or 7.0, 7.8 and 9.0 respectively.
The model used to estimate the effect of fuel RVP on
refueling emissions was developed primarily for the Draft RIA
in support of onboard refueling emission standards.[1] It had
been published prior to its use in the Draft RIA for gasoline
volatility control and received substantial comment and
reanalysis.[2][3] No comments were received with sufficient
technical support to warrant changing this model here. Thus,
it will not be discussed any further.
-------
2-2
II. Summary and Analysis of Comments
In general comments on the evaporative and exhaust
emissions models contained in the DRIA fell into the following
five categories: l) accounting for weathering of in-use fuel,
2) the representativeness of temperature, RVP and tank fill
level inputs used to calculate emission factors, 3) the
validity of the UDI correlations developed for predicting
controlled diurnal emissions, 4) the accuracy of EPA's emission
factor testing and 5) the magnitude of the effect of ethanol
blends on evaporative emissions. Comments in each of those
areas are examined in more detail below.
A. Fuel Weathering
Comments submitted by Texaco, General Motors, and API
maintained that the effects of weathering were not considered
in the DRIA vehicle evaporative emission analysis. EPA stated
in the DRIA that fuel weathering was an important factor in
assessing vehicle evaporative emissions and, in fact, the
decrease in vehicle fuel RVP as a result of the weathering
phenomenon was taken into account in the vehicle evaporative
emission analysis.
The DRIA considered two basic categories of evaporative
emissions generation: 1) hot-soak losses and 2) diurnal
losses. As explained on page 2-34 of the DRIA, fuel weathering
was not included in the hot-soak emissions analysis because
neither ambient temperature effects nor the effect of
temperature and RVP variability have been included in the
hot-soak model. It is also explained that including these
effects would largely offset the predicted decrease in hot-soak
emissions due to fuel weathering. No evidence was presented
demonstrating that these effects did not or could not offset
each other.
The effects of fuel weathering on vehicle duirnal
evaporative emissions were described in the DRIA on pages 2-71
through 2-89. It is concluded in this section that the effect
of weathering and the variability in in-use fuel tank levels
can be incorporated into the average fleet evaporative in-use
fuel tank fill level. The analysis presented in the DRIA
calculates a design value day in-use fuel tank level of 61.5%
which accounts for fuel weathering, RVP variability and average
in-use fuel tank level. Similarly the value of 53.1% in-use
fuel tank level was calculated for July average temperature
emissions modeling which similarly incorporates the effects of
fuel weathering, RVP variability, average in-use fuel tank
level, as well as July temperature variability. These values
were used, as stated in the DRIA, for the city-specific MOBILES
-------
2-3
environmental analysis runs. Without the consideration of
weathering, these fuel tank levels would have been much lower
and emissions much higher.
B. Emission Factor Modeling Inputs
In general comments related to the representativeness of
the data used to calculate evaporative emission factors fell
into three categories: 1) temperature data, 2) RVP values and
3) in-use fuel tank fill levels. Texaco objected to the use of
a single design day temperature for estimating emissions.
While EPA agrees that use of additional high-ozone day
temperatures would improve the analysis, such information is
difficult to gather for 61 cities each year as the design value
day is changed. Also, EPA's analysis does not focus on any one
city, but on all 61 cities, so the variation in the "average"
of the 61 temperatures should be guite small regardless of
whether 1, 2 or 3 days were used for each city. No commenter
presented a suitable alternative to the inputs used in the DRIA
analysis.
Commenters also stated that EPA should not have increased
each area's RVP to the ASTM maximum limit if it was
historically below that limit. They stated that the fuel
distribution system was such that lower limits in adjacent
states often-kept fuel in a given state below its. maximum level.
EPA's reasoning in raising the levels to the ASTM limits
was that levels have been rising nationwide over time as newer
vehicles were designed to handle higher and higher fuel RVP.
In fact, since some areas are already above their ASTM limits,
the use of these limits could be considered conservative.
Also, the economic incentive to add butane to gasoline is the
same at high or low RVP. Only vehicle performance, which is
also a function of climate, vehicle model, and maintenance
level, holds RVP in check. The role of high RVP in hindering
vehicle performance is often masked by these other factors.
On the other hand, EPA now has two additional years' of
MVMA summertime RVP surveys (1986 and 1987) and levels have not
been rising. Thus, at least for this analysis of interim RVP
control, 1987 RVP levels will be assumed to hold constant in
the future absent RVP controls. These revised baseline RVPs
are shown in Table 2-1.
C. UDI correlation
Texaco and API pointed out. that the UDI equations for
predicting duirnal emissions were derived from an EPA data base
limited to three RVP test conditions. While this data base is
limited with respect to the number of test conditions, the EPA
-------
2-4
Table 2-1
In-Use Fuel Volatility (RVP) Data
MSA/CMSA Name In-Use RVP* (psi) 1987
Boston MA 10.8
Greater Connecticut 10.8
New Bedford MA 10.8
Portland ME 10.8
Portsmouth-Dover-Rochester NH-ME . 10.8
Providence RI 10.8
Springfield MA 12.8
Worcester MA 10.8
Atlantic City NJ 11.4
New York NY 11.2
Vineland-Millville-Bridgeton NJ 11.4
Allentown-Bethlehem PA-NJ 11.4
Baltimore MD 11.0
Erie PA 11.7
Harrisburg-Lebanon-Carlisle PA 11.4
Lancaster PA 11.4
Philadelphia PA-NJ 11.4
Pittsburgh PA 11.7
Reading PA 11.4
Richmond-Petersburg VA 11.0
Scranton-Wilkes Barre PA 12.2
Washington DC-MD-VA 11.0
York PA 11.4
Atlanta GA 10.7
Birmingham AL 11.5
Charlotte-Gastonia-Rockhill NC-SC 10.7
Chattanooga TN-GA 10.7
Huntington-Ashland WV-KY-OH 11.0
Louisville KY-IN 11.0
Memphis TN-AR-MS 10.7
Miami-Hialeah FL 10.3
Nashville TN 11.5
Tampa-St.Petersburg-Clearwater FL 10.3
In-Use RVP associated with ozone design value day.
-------
2-5
Table 2-1
(continued)
In-Use Fuel Volatility (RVP) Data
MSA/CMSA Name
Akron OH
Canton OH
Chicago IL
Cincinnati OH-KY-IN
Cleveland OH
Dayton-Springfield OH
Detroit MI
Grand Rapids MI
Indianapolis IN
Milwaukee WI
Muskegon MI
Baton Rouge LA
Beaumont-Port Arthur TX
Brazoria TX
Dallas-Ft.Worth TX
121 Paso TX
Galveston-Texas City TX
Houston TX
Lake Charles LA
Long View-Marshall TX
New Orleans LA
San Antonio TX
Tulsa OK
Kansas City MO-KS
!3t. Louis MO-IL i
Denver-Boulder CO**
Salt Lake City-Ogden UT**
Phoenix AZ
In-Use RVP* (psi) 1987
11.7
11.7
11.6
11.7
11.7
11.7
11.5
11.6
11.6
11.6
11.6
10.7
9.9
9.9
9.9
9.0
10.7
10.7
10.7
9.9
11.4
10.3
9.8
9.8
10.1
9.7
9.7
8.5
**
In-Use RVP associated with ozone design value day.
High-altitude area.
-------
2-6
emission factor data base contains hundreds of vehicle tests,
by far the largest of its kind, representing the broadest range
of vehicle types, model years, etc. Also, it is important to
note that additional data at 27 test conditions obtained in the
ATL study described in the DRIA follows the same trends as the
EPA date. This additional data is presented in Figure 2-10 of
the DRIA and supports the best fit polynomial equations derived
from the EPA data. In fact, use of the emission averages of
the ATL data at the three EPA test conditions does a very good
job of predicting all 27 data points. Thus, absent a better
approach, EPA will continue to use the UDI model based on EPA
test conditions.
As an alternative to EPA's model, a number of oil and auto
companies submitted, and recommended that EPA use, a model
recently developed by Radian, Inc. under the auspices of the
the Coordinating Research Council. This model contains a
number of features not able to be considered in the Draft RIA
model, like the amount of each day's diurnal temperature rise
idle vehicles tend to experience. However, the model at this
point in its development contains a number of major
disadvantages relative to EPA's which argue against its use at
this time.
First, and foremost, the model is entirely statistical,
consisting of correlations of data versus test parameters.
This prevents its use outside of the test conditions, since no
engineering model is used to demonstrate the validity of
extrapolation. Unfortunately, the range of summer climates
occurring in the U.S. is much wider than the range of test data
and extrapolations must be performed.
Second, some of Radion's statistical techniques result in
the consistent underestimation of measured emissions even at
the test conditions. This arbitrarily leads to a reduction in
the estimated effect of RVP on emissions, which is unacceptable.
Third, the most apparent advantage of the Radian model in
estimating partial diurnal may not be as strong as it initially
appears. Diurnal emissions are a function of available
canister capacity, as well as fuel tank emissions to the
canister. Recent modelling by EPA of the interaction of tank
vapors, canister loadings, and purge rates to the engine
indicates that canisters are likely full at the end of the
day. [4] While more work needs to be done in this area, EPA's
running loss testing confirms that many vehicles' canisters are
fully loaded at today's RVPs on high temperature days typical
of high ozone conditions.[5] Thus, at least at the initial
conditions of this interim RVP control analysis, EPA's emission
factor test conditions of a complete diurnal with a purged
-------
2-7
canister is almost certainly more accurate than Radian's
partial diurnals with a purged canister; reality being a
combination of partial and full diurnals with a full canister.
D. Accuracy of EPA Emissions Modeling Testing
API and Sun Refining and Marketing Company expressed
concern over the representativeness of the data base used for
omissions modeling. Specifically fuel blending and the
accuracy of measurement of tank fuel temperatures during test
procedures were questioned.
With respect to fuel blending Sun Oil presented test data
for a single vehicle showing that it was the difference in the
high-end volatility (T90) of EPA's 9 and 11.5 RVP test fuels,
and not RVP, that caused the exhaust emission effect. In
contrast, EPA's own comprehensive testing of a vehicle showed
that it was not any difference in the liguid fuel, but tank
vapors generated during the test, which caused the effect.[6]
Recent running loss test results indicate that exhaust
emissions are strongly affected by tank vapors, much more than
EPA's emissions factor testing would indicate.[7] Thus, even
:.f part of the effect seen with the emission factor testing is
due to differences in Tgg, the whole effect estimated using
EPA's emission factor testing likely underestimates the effect
i.n-use. Thus, the DRIA analysis will continue to be used.
E. Ethanol Effects on Evaporative Emissions
The Ad Hoc Ethanol Committee submitted technical data
describing the average volatility increase for ethanol blended
fuels. Their data resulted in a 0.76 psi RVP increase for a.
10% ethanol blend in gasoline rather than the 1 psi increase
used by the EPA for its analysis. Further data submitted by
API also supports this 0.76 psi RVP increase at 11.5 gasoline
RVP.[8] However, API's model is more realistic in that it
estimates an RVP increase which depends on the base fuel RVP.
Thus, since the two models match at their commonality, but the
API correlation is more flexible, the API correlation is
preferable. However, since a temporary allowance of 1 psi is
being granted to ethanol blend, until the final analysis of the
second level of RVP control, this factor is not an issue here.
III. Final Analysis
The analysis of comments presented above indicates that
EPA's DRIA emission factor models are still the most
expropriate to use. EPA is continuing its development of these
models and improvements will be forthcoming with the release of
-------
2-8
MOBILE4 later in 1989. However, the modifications envisioned
lead to an increase in the emissions effect of RVP. A
preliminary analysis of these revisions shows this increase
could be quite substantial.[9] Thus, continued use of the DRIA
would be very conservative. As shown in Chapter 5, use of the
DRIA estimates still results in a very cost effective interim
program. Thus, for simplicity, the continued use of the DRIA
estimates is sufficient. The preliminary inclusion of running
losses, etc., referenced above, will only be used to indicate
sensitivity and direction of effect. Use of revised emission
factor models will be reserved for the final analysis of the
second level of RVP control.
-------
2-9
Reference - Chapter 2
1. Draft Regulatory Impact Analysis: Proposed
Refueling Emission Regulations for Gasoline-Fueled Motor
Vehicles, Volume I, Analysis of Gasoline Marketing Regulatory
Strategies, U.S. EPA, OAR, July 1987.
2. Evaluation of Air Pollution Regulatory Strategies
for Gasoline Marketing Industry, U.S. Environmental Protection
Agency, Office of Air and Radiation, Office of Air Quality
Planning and Standards, and Office of Mobile Sources,
E]?A-450/3-84-012a (Executive Summary - EPA-450/3-84-012b).
July 1984.
3. Evaluation of Air Pollution Regulatory Strategies
for Gasoline Marketing Industry — Response to Public Comments,
U.S. Environmental Protection Agency, Office of Air and
Radiation, Office of Air Quality Planning and Standards, and
Office of Mobile Sources, EPA-450/3-84-012c, July 1987.
4. David B. Bartus, PT Evaporative Emissions Model,
Description and Users Guide, U.S. EPA, OAR, QMS, ECTD, SDSB,
September 1988.
5. Running Loss Test Program: Interim Results, U.S.
EPA, OAR, QMS, ECTD, SDSB, September 16, 1988.
6. Alan E. Schuler, Effects of Gasoline Volatility on
th.e Hydrocarbon Exhaust Emissions From a 1984 Oldsmobile
Cutlass, U.S. EPA, OAR, OMS, ECTD, SDSB, August 1987.
7. Exhaust Emissions on Repeated LA-4s, Running Loss
Test Program, Data Extracted from MICRO Data Base, January 18,
1989.
8. Letter from Dr. Terry F. Yosie, API, to Mr. Charles
L. Gray, Jr., June 28, 1988.
9. Memorandum to The Administrator, The Effect of
Vehicle Running Losses on Future Ozone Non-Attainment, from Don
Clay, Acting Assistant Administrator, October 6, 1988.
-------
Chapter 3
Environmental Impact
This chapter examines the environmental impact associated
with the interim control of RVP. It discusses the control of
VOC emissions in order to control tropospheric ozone levels;
including discussions of the need for ozone control, the
expected emission reduction effect of these controls, and the
impact of the controls on ozone levels.
I. Need For Ozone Control
A. Synopsis of NPRM
Ozone is a powerful oxidant which is formed in the
troposphere by photochemical reactions of volatile organic
compounds (VOC) and oxides of nitrogen (NOx). Ozone affects
humans by irritating the respiratory system and reducing lung
function. Laboratory studies suggest that it also may actually
damage lung and other tissues. This damage may impair
breathing and reduce immunity to disease for people in good
health, and may be even more severe for people with
pre-existing respiratory diseases. In plants, oxidation by
ozone can impair tissue function and reduce the yield of some
crops. Oxidation by ozone may also damage materials such as
rubber products.
Although various. HC . controls have, already been
implemented, many areas of the nation continue to violate the
ozone NAAQS. Based on the three-year period of 1982-84, EPA
determined that 73 urban areas were exceeding the ozone
standard. (Twelve of these areas are located in California.)
In order to determine the need for future hydrocarbon
control, EPA looked not only at the present state of ozone air
quality, but also at projected future air quality trends.
Estimates of future air quality were made for the 61
non-California urban non-attainment areas. (Since California
has separate motor vehicle standards, California non-attainment
areas were not modeled.) Current attainment areas were
excluded from the future air quality projections even though
some of the areas were close to the standard and may have been
projected to become a non-attainment area in the future. The
air quality modeling relied on certain assumptions regarding
emission rates, growth rates, control technologies, emission
standards, and control efficiencies that are described in more
detail in the following sections. Based on these analyses of
future air quality, EPA projected that there will likely be
improvements in air quality from 1988 to 1995 due to the effect
of current emission standards for mobile and stationary
-------
3-2
sources, but that most of the large urban areas modeled remain
in non-attainment throughout this period. However, without
further controls, growth is projected to offset these
improvements and to cause the problem to worsen after 1995.
Based on estimates made by the EPA, VOC emission reductions of
50 to 80 percent appeared to be necessary to bring some urban
areas into compliance.
Both the proposed volatility regulation and the proposed
refueling regulation result in emission reductions in both
non-attainment and attainment areas. The need for control in
non-attainment areas is evident. It is also, however,
beneficial to control volatility (and hence, VOC emissions) in
attainment areas. Two key ozone benefits were noted in the
NPRMs. Because VOC emissions from one area may be transported
through the atmosphere to another area, emissions in attainment
areas can add to the problems in non-attainment areas.
Therefore, controlling emissions in attainment areas will also
help non-attainment areas to reach compliance.
B. Summary and Analysis of Comments
The comments received concerning the need for ozone
control addressed many different aspects of this issue. These
many aspects, however, fall into three main areas: 1) the
health effects of ozone exposures, 2) the extent of the ozone
problem, and 3) the more general issue of the type of control
needed.
1. Ozone Health Effects
Comments on health effects varied greatly. NRDC stated
that at concentrations of 50 percent of the current ozone
NAAQS, suppression of the immune system has been observed in
laboratory animals. The American Lung Association (ALA) also
agreed that the current ozone standard does not adequately
protect public health. ALA commented that clinical studies
have shown that adverse respiratory health effects result from
experimental ozone exposure at the current standard level.
NESCAUM stated that ozone pollution is one of the most serious
and widespread public health problems in the Northeast U.S. It
also stated that recent health data strongly suggests that the
existing ozone NAAQS may not be strict enough to protect public
health. It also claims that much terrestrial damage has
occurred due to ozone exposure at ambient concentrations well
below the current health standard.
Others commented, though, that any health effects are very
short term in nature and that those showing health effects
recover quickly. API commented that the line between
attainment and non-attainment seems arbitrary and that it is
-------
3-3
improbable that any real public health consequences are
associated with the transition from non-attainment to
attainment.
Commenting on the margin of safety required by the Clean
Air Act, Chrysler felt that it is irrelevant whether or not an
area is projected to be in borderline compliance, as long as
they reach attainment. It felt that there already exists a
significant safety margin built into the current standard to
protect public health. Therefore, it argued, as long as
compliance is met, no extra margin of safety is needed. NRDC,
however, stated that recent scientific evidence shows that the
ozone NAAQS will have to be tightened in order to protect the
public health with an adequate margin of safety.
These comments demonstrate that additional information may
be needed to resolve concerns about the appropriate level of
the ozone standard. This rulemaking, however, is not the
proper forum to attempt such a resolution. The Agency is
already investigating this issue in its periodic (5-year cycle)
review of each NAAQS. Should it be determined that
modifications to the current ozone standard are necessary, they
will be made in a separate rulemaking. At this time, control
decisions have to be based on the current standard.
In response to Chrysler's comment, it is true that the
Clean Air Act requires an adequate margin of safety to protect
the public health. Thus, EPA does not base its regulatory
decisions on areas that are in "borderline attainment". This
does not, however, mean that the extent of an area's compliance
is irrelevant. Concerns such as NRDC's on the margin of
safety, as well as concerns about the accuracy of ozone
projections, make the projection of an area to be in borderline
attainment very significant.
2. Extent of Ozone Problem
Regardless of the level of ozone at which an unhealthy
environment exists, the extent of the ozone problem which
exists in the nation is also under debate. Phillips Petroleum
commented that few urban areas frequently or significantly
exceed the ozone standard. It also commented that, with the
exception of Southern California, the ozone problem is largely
unnoticed by the general public since exceedances of the ozone
standard are so few. Other commenters also noted that on an
hour-by-hour basis, most areas are in compliance with the ozone
standard more than 99 percent of the time. API agreed with the
I2PA that there is an ozone problem, however, it felt this
problem has been overstated by the EPA. Sun Oil also felt that
EPA has overstated the problem. They believe that, except for
the Los Angeles area, most non-attainment areas are in
compliance with the ozone standard during most of the summer
ozone season. MVMA commented that the EPA has already
-------
3-4
determined through its NAAQS process that an ozone
concentration of 0.12 ppm or less is by definition adequate to
protect public health and welfare from any adverse effects.
Many comments stated that the EPA has over-estimated :he
current ozone problem because the method for determ: ing
whether or not an area was in attainment was erroneous. .hey
feel that the method does not represent a quantitative vv v of
ozone exposure and it overstates exposure duration and ..otal
population exposure. Sohio and API both commented that
determining attainment or non-attainment with a single monitor
is unrealistic and overstates the problem. Sohio feels that
the average of all monitors in the area may be more
representative of typical ozone season air quality. API and
Sohio stated that overall ozone levels are well below the
standard, but that the current method of reporting ozone
non-attainment data does not show that. Sohio also feels that
since the design value is based on a reading from a single
monitor, it does not show the progress achieved in reducing
ambient ozone levels, and, therefore, does not measure the
extent of non-attainment. Sohio commented that monitors with
exceedances may not be in highly populated areas. It said that
in most non-attainment areas, the population living or working
in the area is much lower than the total population. API and
Sohio also stated that under the current method, one year's
data, even if it is uncharacteristic, is enough to keep an area
out of attainment for a three-year period. One last commenter
questioned whether or not a one-hour standard for ozone was
appropriate. They believe that a more practical long-term
solution would be going to an eight or twenty-four hour average
standard.
EPA does not deny that ozone levels in urban areas are
generally below the level of the standard (0.125 ppm) the
majority of the time, nor does it claim that its monitoring
system reports an ozone level to which the entire area is
exposed. Nevertheless, it does maintain that its current
method of determining air quality, with respect to ozone
levels, is reasonable. This is because the Agency considered
these issues in setting the level of the standard. It was
decided that one-hour ozone concentrations greater than 0.125
ppm, in any part of an area, more than once per year would be
indicative of unacceptable air quality. EPA will consider
comments on modifying the nature of the ozone standard at the
appropriate time in its review of the ozone NAAQS. However, as
noted before, such modifications are clearly beyond the scope
of this rulemaking.
3. Necessary Control Programs
Many commenters stated that because many areas in the
nation still have not reached attainment of the ozone standard,
further control of HC emissions is necessary. Some commenters
-------
3-5
argued for volatility controls, some argued for refueling
controls, and some argued for both. The Texas Air Control
Board felt that onboard controls would be a very cost effective
strategy, as did API and many other commenters. API, however,
does note that before forcing costly hydrocarbon controls, it
would be useful to have a better understanding of the role of
NOx, the role of transport, and the overall hydrocarbon
inventory. It cited the "National Acid Precipitation Report",
stating that two-thirds of VOC emissions east of the
Mississippi and three-fourths of VOC emissions west of the
Mississippi are from natural sources. The Conservation Law
Foundation of New England (CLF) agreed with EPA's assumption
behind the proposed volatility rule that vehicle-based controls
alone are not enough to address the short-term ozone
non-attainment problem. This is because emission reductions
firom such controls only occur after a long period of time
needed for fleet turnover. It also felt that the EPA was
correct in realizing that emission reductions beyond those
available at zero cost are necessary to achieve short-term air
quality improvement. It felt that the in-use fuel volatility
restrictions are technically feasible and cost effective, and
thus should be adopted. NRDC and the American Lung Association
felt that HC reductions beyond those currently required in
non-attainment areas are necessary to protect public health.
They felt that volatility, onboard, and Stage II controls are
all necessary to achieve attainment. NRDC stated that $2000
per ton should not be a cost-effectiveness ceiling for ozone or
carcinogen controls. NESCAUM, in commenting on the ozone
problem in the Northeast, stated that long-term attainment of
the ozone standard can only be reached through a regionwide
program based on all available control strategies. It also
felt that a reason for the failure to attain the ozone standard
in the Northeast has been an incomplete understanding of the
complex process of ozone formation. This would lead some
states to underestimate the required reductions to meet the
ozone standard, and to overestimate the ozone reductions
obtained from various control measures.
Some commenters, however, felt that there is not a problem
of ever-worsening air quality. Therefore, they claim that no
nationwide "crash" program is needed to reduce emissions.
Toyota felt that the effectiveness of current control programs
is being underestimated by EPA.
Clearly there is disagreement as to whether the proposed
controls are necessary. However, none of these comments
provided sufficient rationale to change EPA position on VOC
control. The Agency remains convinced that cost effective VOC
controls are the most appropriate ozone strategy at this time.
(It should be noted that a $2000 per ton cost-effectiveness
ceiling, as NRDC implied, does not exist in EPA's ozone
policy.) Thus, the argument presented by the automotive
industry that refueling emissions should not be controlled
-------
3-6
since they only account for two percent of the VOC inventory is
not valid; size of the reduction is not important, cost
effectiveness is what matters, since many smaller benefits ^an
account for large reductions when considered together. In
regard to DOE's comment, it should be noted that EPA lid
consider both short-term and long-term costs and bene.-ts
before proposing these controls.
It is true that natural VOC emission can play an impor-tant
role in ozone formation, and in fact, EPA is investigating this
issue. The other issues such as the role of NOx and of
transport are also important and should be investigated
further. Nevertheless, this does not alter the fact that
control of VOC emissions such as gasoline vapors is an
effective means of reducing ozone levels.
II. Emission Inventories
A. Synopsis of NPRM Analysis
In order to estimate the environmental impacts of the
control options, a computer model was first used to calculate
VOC emission factors in grams per mile. The emission factors
were calculated for exhaust, evaporative, and refueling,
hydrocarbon (HC) emissions, and were multiplied by VMT
estimates to obtain inventory projections of- future VOC
emissions. The inventory projections in turn were used to
determine the emission reductions from the various control
options.
The version of MOBILES used in the DRIA analysis was
different than the released version. In order to incorporate
RVP effects and refueling emissions, the analyses used an
in-house version of MOBILES (hereafter refered to as
MOBILE3.9). Since evaporative emissions are dependent on
temperature, the in-house version also allowed for the use of
climatic data reflecting actual temperatures in the areas
modeled. This model was used for both the RVP and refueling
analyses. The analyses differed, however, in that the RVP
analyses incorporated city-specific temperatures and RVPs,
while the refueling analyses did not.
In the RVP analyses, two different types of city-specific
temperatures were used for the different analyses performed.
Environmental impacts were modeled using temperatures from the
design value day of the area. (The design value is defined to
be the fourth highest one-hour ambient ozone concentration
occuring in a three year period.) Inventory projections based
on the design value day temperatures were used to compare and
rank various control programs.
The economic impact of RVP control was determined using
July average temperatures rather than design value day
-------
3-7
nperatures. EPA used this average temperature approach
::ause the economic benefits of recovering evaporative
issions occur throughout the summer, not just on high ozone
/s. In particular, the average temperatures during the month
July were used because they correspond to the in-use fuel
latility survey data used in this study. As with the design
lue day temperatures, July average temperatures were used to
Iculate emission inventories and emission reductions in
ojection years under various control options. These
ventories wore then used to estimate economic credits (i.e.,
a evaporative recovery credit and the attainment area VOC
ntrol credit).
City-specific alcohol-blend market shares were also used
the modeling in order to obtain more accurate emission
ctors by accounting for the fraction of alcohol blend fuel in
e market. EPA estimated that alcohol blends have a 1.0 psi
gher RVP than the base gasoline. The market share data was
so used to evaluate the impact of various alternative
eatments of alcohol blends. In the modeling, it was assumed
.at the RVP of gasohol was 1.2 psi greater than that of
soline (i.e., 1.0 psi RVP effect with a 0.2 psi average
mmingling effect). In those scenarios where alcohol blends
re required to meet the same RVP restrictions as gasoline,
.e in-use RVP of alcohol blends was assumed to be 0.2 psi
eater than gasoline, due solely to commingling.
In addition to city-specific temperatures and' alcohol
.end market shares, city-specific in-use fuel volatility
irvey data were used in the RVP modeling. The 1983 RVP data
'presented average volatilities for non-alcohol-containing
ileaded gasolines in the summer months. These RVP levels were
;ed to develop the 1983 base year inventories. The 1985
irvey data were used in all the projection years. For these
)85 values, if the area surveyed had an RVP less than the ASTM
Lmit for that area, the ASTM limit was used in place of the
irveyed RVP level. This was done based on the assumption that
ae RVP in the area would continue to rise until it reached the
3TM limit. If no RVP data was available for a given
jn-attainment area, the RVP of the nearest survey area was
sed. Also, the most appropriate RVP value would be that for
ie month during which the design value day occurred. Since
\e surveys were performed only in July, if the design value
ly was not in July, the RVP was adjusted to represent the
orrect ASTM class for the month of the design value day.
The above city-specific input were incorporated into the
3BILE3.9 model which then calculated HC emission factors for
ach of the urban, non-California ozone non-attainment areas
odeled. These non-attainment area emission factors were then
onverted into nationwide emission factors based on a
opulation-weighted average of the individual emission
actors. FLeetwide exhaust, evaporative, and refueling
-------
3-8
emission factors were calculated for each gasoline-fueled
vehicle category (LDV, LDT, HDV) . The HC emission facto.rs for
gasoline-fueled vehicles in the RVP analyses were calculated
assuming both that onboard refueling emission controls would be
in effect as of the 1990 model year and also assuming no
onboard controls.
In estimating refueling emissions, because all the
necessary city-specific inputs were not available, MOBILES.9
was used to develop nationwide emission factors directly, using
national average inputs such as fuel volatility and
temperatures. The analysis also accounted for tampering with
the evaporative emission control systems, which may occur in
use. The types of tampering identified as those which could
affect onboard refueling controls were• disconnection and
removal of the evaporative emissions storage canister. This
was assumed to occur at rates equal to those of current control
systems. As in the RVP analysis, fleetwide exhaust,
evaporative, and refueling hydrocarbon emission factors were
calculated for each gasoline-fueled vehicle category. It
should be noted, however, that these were calculated assuming
no RVP control.
Using the calculated nationwide emission factors,
inventory projections of future VOC emissions were
estimated. Non-California, urban, non-attainment area
inventories were calculated for both mobile and stationary
sources, both as a whole and subdivided into several individual
source categories using assumptions concerning growth rates and
future technological improvements.
These projections were used to generate projected emission
reductions associated with each control strategy. The
nationwide emission reductions were used in ranking
control-effectiveness and cost-effectiveness of the RVP control
programs. The non-attainment area inventories were used as
input for the ozone air quality modeling and for estimating the
cost effectiveness of RVP controls.
Nationwide inventory projections for the calculation of
economic recovery credits for RVP control were made using
emission factors based on July average temperatures for the
non-attainment areas. For the ozone air quality modeling for
RVP control, the analysis used emission factors based on the
design value day conditions. The design value day was used as
the basis for the emission projection since the ozone model
focuses on that day. For all the RVP projections made, a base
year of 1983 was used with projection years of 1988, 1990,
1992, 1995, 1997, 2000, 2010.
-------
3-9
B. Summary and Analysis of Comments
Some comments were received on the type of temperature
data used in the modeling. MVMA commented that there is no
clear relationship between the design value level and the July
average temperatures used. MVMA feels that better correlation
with temperature could be obtained with an ozone statistic more
robust than the design value. When MVMA repeated the analysis
to calculate inventory projections, it used July minimum and
maximum temperature data whereas EPA used the design value day
minimum and maximum temperature data for each city. It noted
that in some cities this resulted in higher temperatures, while
in others it resulted in lower temperatures.
MVMA is correct in noting that there is no clear
relationship between design value and July average
temperatures. However, this is not surprising since there are
many other city-specific factors that impact the magnitude of
the design value, such as VOC emissions and VOC:NOx ratios. In
regard to the minimum and maximum temperatures used in the
modeling, MVMA presented no arguments that their approach was
more appropriate, but did note that the difference was variable
from city to city. Lacking other information, the design value
day temperatures actually occured and produced high ozone
levels. The July-average temperatures have a much weaker
connection with ozone levels and the effect of emission
controls on ozone levels. Thus, there is no reason for EPA to
alter its use of temperatures in its air quality modeling.
MVMA commented that the growth rates used were not
city-specific, yet should have been, since much of the other
air quality projection data used was city-specific. It also
suggested that growth rates should also be city-specific since
they vary widely from city to city. J.G. Bathe stated that
EPA's growth rates are too high, and lead to overestimated
emissions.
EPA is, in fact, currently working to develop improved
growth rate assumptions. This work includes consideration of
city-specific aspects. Unfortunately though, this work is as
yet incomplete, and thus not available for this rulemaking.
Also, none of the commenters presented any information from
which growth rates for each city could be developed, nor
demonstrated that updated rates would be different. Even
without any future growth, due to the magnitude of the current
ozone problem, the conclusions drawn from the air quality
analysis would not be substantively altered. Thus, for this
analysis of near-term RVP control, EPA will use the same growth
rates as were used in the NPRM analyses.
The choice of 1983 as the base year for the analysis was
thought by some to introduce significant error in the results.
Sun Oil stated that the summer of 1983 was unusually warm and
-------
3-10
had an unusually large number of ozone exceedances. It felt
the EPA must develop statistical attainment data using normal
temperature information determined over an extended period of
time (15-30 years). MVMA also commented that EPA selected an
unrepresentative base year design level on which to base the
air quality calculations.
While it is true that 1983 was among the warmest in recent
years, this was not the reason EPA chose it. Actually, at the
time the DRIA analysis was performed, the most recent
three-year period of design values was 1982-1984, with 1983
being the middle, or base year. Also, it is not obvious that
the high temperatures of 1983 would make it inappropriate for a
base year. Furthermore, recent preliminary ozone 'data
indicates that the summer of 1988 was even warmer and had an
even more severe ozone problem.
MVMA objected to some of the adjustments made to the NEDS
inventory for the base year 1983. It feels that the
adjustments tended to decrease the contribution of all
stationary sources to total NMHC emissions, and thus to
increase the apparent importance of motor vehicles. It feels
that this could not be validated with the current knowledge of
emissions of stationary sources. MVMA also disagreed with the
resulting trends in the inventory.
While it may be true- that improvements could be made to
the estimation of stationary source emissions, such
improvements would not substantively change the conclusions of
this analysis. The acceptability of these regulations will be
judged based on the cost effectiveness of the reduction of
mobile source emissions, and thus the estimation of stationary
sources is not significant.
In the reanalysis which MVMA had done, they used more
city-specific data than EPA, which they say is preferable.
Whereas EPA used a 1983 implementation date for I/M programs
for all cities, MVMA used city-specific implementation dates.
MVMA also accounted for Stage II controls being implemented in
New York, New Jersey and St. Louis, and used city-specific
model year start dates and vehicle classes.
In past analyses, EPA has noted that its modeling was
performed on a nationwide basis, and that as such, it was not
appropriate to predict the exact impact of rules on each
particular area. Nevertheless, this approach was sufficient
given the nature of EPA's past mobile source controls.
However, control of in-use RVP required a more city-specific
approach than used previously. For this program, it was
essential that EPA's model be able to account for city-specific
RVPs and temperatures. Other inputs, such as for I/M programs,
are not considered on a city-specific basis because the models
are not capable of handling them at the present time. This
-------
3-11
approach is expected to be sufficiently accurate given the
neiture of the Agency's modeling goal: to provide an estimate
of the impact of control on a nationwide basis. In regard to
MVMA's Stage II comment, it applies primarily to the onboard
refueling analysis and not the RVP analysis.
Some commenters questioned the number of non-attainment
areas that should be included in the analysis. MBS feels that
only those areas expected to be in non-attainment in the future
should be included. OMB went further to say that the effect of
compliance due to volatility controls should also be considered.
It has been Agency policy to include all current
non-attainment areas in air quality analyses. To ensure
consistency it is necessary to continue this practice. Also,-
analyses based on future years would be very dependent upon the
year selected, which could lead to significant policy
problems. However, the number of non-attainment areas is
expected to be very similar in either case without additional
controls. Thus any change in this policy would not affect the
conclusions drawn here.
MBS also noted that their analysis used a 30-year model of
the fleet, instead of a 20-year model .like EPA used. They feel
that this is appropriate because vehicles more than 20 years
old contribute significantly to total fuel consumption.
Chrysler stated that MVMA statistics show that fleet turnover
is slower than that used in the NPRM (e.g., it takes 15 years,
not 13 years, for 90 percent of cars and trucks to be replaced).
While it is true that a 30-year model is likely to be more
accurate than a 20-year model, the difference would not be
expected to be substantial. EPA's model accounts for all
model-years more than 19 years old on an aggregate basis. This
is reasonable since such vehicles account for less than one
percent of vehicle-miles travelled (VMT). Since Chrysler did
not include information about the model years involved in the
MVMA study showing 90 percent of vehicles being replaced after
15 years, it is not possible to resolve the difference from
EPA's estimate. It is likely that MVMA estimate involved
different model year than EPA's, which could explain the
difference since vehicle sales vary from year to year due to
economic influences.
C. Final Analysis
For the 1989 through 1991 timeframe, EPA will require
gasoline sold during the summer months to have its volatility
reduced below the maximum ASTM limits to 10.5 psi (primarily in
Cliiss C areas), 9.5 (primarily in Class B areas), and 9.0
(primarily in Class A areas, no reduction). (The specific
standards for each area are outlined in Chapter 1.) The
-------
3-12
analysis focuses on 1990 which is representative of the entire
1989-1991 control period.
The analysis of comments presented in Chapter 2 indicates
that EPA's DRIA emission factor models are still the most
appropriate to use. EPA is continuing its development of these
models and improvements will be forthcoming with the release of
MOBILE4 later in 1989. However, the modifications envisioned
lead to an increase in the emissions effect of RVP. Thus,
continued use of the DRIA would be very conservative. As shown
in Chapter 5, use of the DRIA estimates still results in a very
cost effective interim program. Thus, for simplicity, the
emission inventories found in Chapter 3 of the DRIA for the
case without prior onboard implementation will be used here to
estimate the emission reductions resulting from volatility
control. Unlike in the DRIA, however, estimates are being made
on an ASTM class-specific basis. As a result, ASTM class
specific emission reductions were derived by assuming that
control to 9.5 in Class B was equivalent to 10.9 in Class C,
and 9.0 in Class A equivalent to 11.5 in Class C. In addition,
uncontrolled fuel is no longer considered to have an RVP of
11.7 psi, but rather to be 11.3 psi in Class C areas, 10.0
(11.5 equivalent) in Class B areas, and 8.6 (11.0 equivalent)
in Class A areas (as determined in Chapter 4). This change
caused emission reductions to decrease substantially below the
DRIA estimates, and in fact caused emission reductions in-Class
A areas to disappear entirely.
One additional change from the methodology of the DRIA was
the value used for 1990 nationwide VMT. Comments discussed in
Chapter 4 stated that by using the MOBILES fuel consumption
estimates EPA had significantly underestimated the refining
costs. As a result, in Chapter 4, new refining cost estimates
are based on DOE fuel consumption estimates.[1] In order to
remain consistent with this change, 1990 nationwide VMT also
had to be changed to reflect the DOE estimates. This was
incorporated merely by multiplying the emission reduction
estimates from the DRIA by the ratio of 1990 nationwide VMT as
estimated by DOE (approximately 2068.57 billion miles), to that
as used in the NPRM (1760.81 billion miles).
In order to determine the emission reductions in each ASTM
class individually, the emission reductions were multiplied by
the fraction of VMT (assumed to be the same as the fraction of
fuel sales) which occurs in each ASTM region (4.02% in Class A,
34.19% in Class B, and 61.79% in Class C) . The resulting VOC
emission reductions are shown in Table 3-1.
As described in Chapter 2, these figures represent very
conservative estimates given more recent knowledge concerning
running losses and excess evaporative emissions. To show the
potential for greater control, to these emission reductions
were added "a recent estimate of excess evaporative and running
-------
3-13
Table 3-1
Nationwide Year Round Emission Reductions
for the Interim RVP Control Program
Design Value Day (1000 ton/yr)
Class A
Class B
Class C
Nationwide
DRIA
Estimate
0.0
208.46
465.30
• 673.76
Preliminary
Running Loss and
Excess Evaporative
Emissions
0.0
413.18
887.62
1300.80
Preliminary
Total
(for Sensitivity
Purposes)
0.0
621.64
1352.92
1974.56
July Average (1000 ton/yr)
Class A
Class B
Class C
Nationwide
DRIA
Estimate
0.0
158.25
360.05
518.30
Preliminary
Running Loss and
Excess Evaporative
Emissions
0.0
313.67
686.83
1000.50
Preliminary
Total
(for Sensitivity
Purposes)
0.0
471.92
1046.88
1518.80
Note: Control period emission reduction estimates are
estimated to be 42.56 percent of year-round reductions
based on the fraction of fuel sold during that period.
-------
3-14
loss emission reductions.[2] Running loss emission factors for
the interim RVP control program were estimated by fitting a
curve through the three data points of running loss emission
factors as a function of fuel RVP, while excess evaporative
emission factors were derived for the scenarios required by
adjusting the estimates at 9 psi RVP by the ratio of ,ie
uncontrolled diurnal index (UDI) at the RVP required to that at
9 psi.[2] The difference between the base case and control
case emission factors was then multiplied by the nationwide "»/MT
estimate for the individual ASTM classes to determine the tctal
excess evaporative and running loss emission reductions. (The
nationwide VMT was estimated from the most recent DOE
information, and then broken into ASTM classes by multiplying
by the fraction of VMT occurring in each ASTM class area.Cl])
July average excess evaporative and running loss emission
reductions were then determined by using the ratio of July
average to design value day emission reductions found in the
DRIA. The resulting emission reductions are shown in Table 3-1.
III. Ozone Modelling
A. Butane and Oxygenate Reactivity
1. Synopsis of the NPRM Analysis
The photochemical reactivity of butane and some oxygenates
(i.e., ethanol and methanol) was discussed in the NPRM.
However, the previously established Agency policy was to treat
all reactive compounds equally. While the rationale for this
position was not discussed in detail in the NPRM, it would
still be useful at this point to review some of the reasons why
EPA has chosen in the past not to consider differential
photochemical reactivity in its control programs. First, there
is the issue of multi-day pollution episodes, of which there
are two types: one where the pollution remains in one area for
a prolonged period due to stagnation, and the other where the
pollution is transported to another area without being
substantially diluted. With both types of multi-day pollution,
the rate at which a particular species reacts becomes less
important since it has a longer time to react. For example, a
slowly reacting compound may not react completely in a single
city, but could be transported to another where it would
continue to react. Consequently, rate data showing that attack
by OH radical occurs slowly, or even single-day modeling
studies showing a compound to be less reactive, are not
sufficient to quantify the effect of a particular compound
during a multiday episode.
Second, it has been shown that the absolute and relative
reactivity of compounds can be affected by changes in HC to NOx
ratios. Specifically, differences in reactivities are smaller
at low HC to NOx ratios. [3] This is important because it is
-------
3-15
expected that the current aggressive HC control strategy will
result in lower HC to NOx ratios, and thus smaller differences
in reactivity.
Third, in many respects the modeling capabilities and
chemical data available are too limited to completely address
this issue. The available models are only approximations of
what happens in a real airshed, and thus the results cannot be
considered to exactly describe the actual effect of any
particular compound. Photochemical modeling requires a great
deal of meteorolgical data to account for things such as
transport and dispersion. Often these data are not completely
available, and assumptions are necessary. Other modeling
simplifications are generally included to reduce computational
time (and thus costs). With respect to chemical data, it is
noted that for complicated molecules there can be too many
reaction pathways to model completely. Even for a relatively
simple molecule like methanol, there has been some question
about its reaction mechanisms.[4] This situation can result in
the exclusion of several mechanisms that are deemed to be
insignificant. Thus predicting the reactivity of a molecule
requires simplifying assumptions, which all add some degree of
uncertainty to any attempted analysis.
Finally, attempts to regulate while considering reactivity.
can easily result in unworkable regulations. As noted before,
future changes, in HC to NOx ratios could change the real-world
reactivity of a compound, which combines with the potential for
changes in modeling technology to make any analyses done at
this point continually subject to change. Also, regulating on
the basis of reactivity could require careful monitoring and
controlling of the chemical composition of fuels. This would
bo much more difficult than controlling a fairly simple
physical parameter such as RVP. Moreover, since fuels tend to
contain a large number of components, the volume of data
necessary to accurately predict the reactivity of some type of
emission (e.g. evaporative emissions) could become
overwhelming. While concerns about the ease of regulating are
not adequate justification for not considering photochemical
reactivities, they do need to be weighed against the potential
for benefits from such consideration. A summary of the NPRM
discussions of butane and alcohol are summarized separately
below.
In the NPRM, the photochemical reactivity of butane was
classified using a 1984 EPA report.[5] This report classifies
compounds as "unreactive," "borderline," or "reactive;"
reactive compounds being those which are significantly more
reactive than ethane (based on smog chamber and/or rate data.)
Using this system butane was classified as reactive. In
support of this conclusion, it was noted that butane would also
be classified as reactive using the GM scale that was developed
in the mid-1960s.[6] The analysis went further to say:
-------
3-16
Modeling and smog chamber data also verify the
contribution of butane to ozone formation in the
troposphere. Based on available information, it is
EPA's position that butane is a photochemically
reactive compound which contributes to ground level
ozone formation. Therefore, reductions in butane
emissions through in-use fuel volatility control are
expected to lead to subsequent reductions in ambient
ozone levels.
Thus, no numerical consideration of the specific reactivity of
butane relative to average VOC was attempted by EPA.
The photochemical reactivities of ethanol and methanol
were discussed briefly in the NPRM; no other oxygenates,
however, were discussed. The only oxygenate which EPA
considered in its air quality analyses was ethanol. For
calculational purposes, ethanol was assumed to be as reactive
as average VOC on a per carbon atom basis, which meant that
ethanol was assumed to be only 62 percent as reactive on a mass
basis. This difference is because ethanol molecules contain
oxygen atoms, and thus ethanol has a much higher mass to carbon
atom ratio than most VOC. This approach was chosen since it is
the oxidation of carbon that contributes to ozone formation in
the troposphere. It was also noted that the rate at which
ethanol reacts with OH radical in the atmosphere (the primary
mechanism for oxidizing VOC) is on the same order as
hydrocarbons such as butane or toluene.
The NPRM discussion also noted that studies had indicated
that methanol is less reactive than typical hydrocarbon
vapors. The studies showed the reactivity of methanol to be
only 2-43 percent as reactive as average VOC on a carbon
basis. These numbers were used to estimate the air quality
impacts of methanol blends.
The increase in emissions of formaldehyde due to
combustion of methanol blends was also discussed. Due to the
very high reactivity of formaldehyde, which offsets the low
reactivity of methanol, the reactivity of exhaust emissions
from methanol blends was assumed to be the same as from
gasoline.
2. Summary and Analysis of Comments
The Agency received many comments (from Sohio, the Ad Hoc
Ethanol Committee, OFA, the Ohio Farm Bureau, NESCAUM, Sun
Refining Co. and GM) which suggested that EPA ought to consider
the photochemical reactivity of evaporative emissions in this
rule. Most of the comments were with regard to butane (and
other light paraffins) and/or oxygenated compounds. The
commenters stated that these compounds are less reactive than
-------
3-17
average VOC. This conclusion was based on rate constants for
reaction with OH radical, the results of smog chamber studies,
and the results of single-day computer modeling studies. Since
light paraffins (and oxygenated compounds for oxygenate blends)
make up a large part of evaporative emissions, the commenters
felt that controlling such emissions will not be as effective
in reducing ambient ozone as EPA has suggested.
Sohio stated that EPA should not rely on smog chamber
studies to determine the reactivity of butane since smog
chambers are not representative of real world conditions.
One of the modeling studies submitted, which was performed
by Systems Applications Inc. (SAI), looked at various control
strategies involving ethanol blends. The model predicted that
even if ethanol blends had higher RVPs than straight gasoline,
they would still result in less ozone production for many
cities. The report explained that this was due to the combined
effect of ethanol's reactivity and its effect on CO emissions
(which contribute to ozone formation). The report went further
to say that there are two key aspects of ethanol's chemistry
that make it less reactive. First, it noted that the initial
reaction of ethanol with hydroxyl radicals generates only half
as much ozone as the initial reactions of typical
hydrocarbons. Second, it stated that acetaldehyde, the
principle intermediate product, is not highly reactive and can
inhibit ozone formation by reacting with NOx to form
peroxyacetylnitrate (PAN).
Similarly, the report stated reasons why the reactivity of
butane is low. It noted the low rate constant for the reaction
of OH radical with butane, which is much lower than the average
value used for VOC in EKMA. It also stated that the main
products of the reaction of OH with butane are acetaldehyde and
methylethyl ketone, and that these products are "not much more
important in further ozone formation than is butane itself."
Some of the comments did address the points noted above,
regarding EPA's rationale for not considering reactivity.
NESCAUM noted that there is generally an incomplete
understanding of the photochemical formation of ozone. Sun
noted that HC to NOx ratios can have a significant effect on
reactivities, and thus that for some areas HC control without
NOx control will not result in attainment. It added that the
mechanisms of transport and dispersal are not well understood.
It also noted that because there are processes which scavenge
ozone in the atmosphere, compounds that produce ozone more
slowly do not allow concentrations of ozone to reach levels as
high as others might. They went further to say that this slow
reaction rate would also allow the compound to be dispersed
before it could produce large amounts of ozone in an urban area.
-------
3-18
Much of the evidence presented to show that butane and
many oxygenated compounds react slowly in the atmosphere was
smog chamber data and" rate constants for reaction with OH
radical. This information, while valid, is insufficient to
allow a quantified estimate of the reactivities of these
compounds in real situations. The qualitative discussions in
the SAI report are interesting; however, some of the points
raised are misleading. First, the report claims that
acetaldehyde is not highly reactive; however, at least one
modeling study has shown acetaldehyde to be more reactive than
formaldehyde, which is accepted as a highly reactive
compound.[4] Moreover, the report notes that acetaldehyde
scavenges NOx by reacting with it to form PAN. This would in
fact be -of some benefit with respect to ozone levels, but there
are also negative aspects of PAN formation: PAN itself is an
air pollutant which causes significant health effects, and it
can be transported long distances before it decomposes and thus
regenerates the NOX. EPA regulates ambient ozone levels as a
surrogate for all oxidants, assuming 'that satisfactory levels
of ozone will result in satisfactory levels of all oxidants.
Substitution of PAN for ozone would clearly be unacceptable.
Another misleading aspect is the comparison of rate constant of
the reaction of OH with butane to the rate constants for
reactions with the theoretical species of the Carbon Bond
Mechanism (CBM) which is used in EKMA. This is not really
appropriate since the butane rate constant represents only the
initial reaction with OH, while the CBM r.ate constants
represents an average of the reactions of all the carbon atoms
initially present as paraffins. Thus the CBM rate constants
also account for the products of the initial reaction, while
the butane rate constant does not.
The corresponding photochemical modeling study is more
useful, yet it still does not resolve the concerns noted above,
especially the concern regarding multi-day pollution. Thus
none of the comments are sufficient to justify reversing
established Agency policy. Also, it should be noted that Sohio
was incorrect in stating that EPA relies solely on smog chamber
studies when considering reactivity.
EPA does not deny that butane and many oxygenated
additives react more slowly than average VOC, or that such
slowly reacting compounds can result in somewhat less ozone
than other compounds. Rather it holds that at this time, it is
not possible to accurately address reactivity issues such that
consideration would be workable and appropriate.
This precedent was reaffirmed just recently in the
rulemaking that established standards for methanol-fueled
vehicles. That rule regulated organic emissions on a carbon
basis, and did not give any allowance for lower reactivity,
even though there is some evidence that methanol has a very low
photochemical reactivity.
-------
3-19
The SAI report submitted by the Ad Hoc Ethanol Committee,
le it does not resolve the concerns noted above, does
jest that the potential reduction of CO from vehicles fueled
a ethanol blends could also impact ozone levels. Since the
nistry of CO is fairly simple and well understood,
itifying the impact of CO reduction does not involve the
3 degree of uncertainty as when dealing with reactivity
afits of other compounds. For this reason, the Agency is
.1 to considering the effect of CO on ozone levels, when
ling with alcohol blends. Unfortunately though, the
ormation cxirrently available is not sufficient to quantify
effect, especially considering the fact that the effect can
7 greatly.
B. Air Quality Projections
1. Synopsis of the NPRMs
The ozone air quality analysis, which predicted future
ne concentrations, was done using the EKMA computer models.
NMOC emission inventories were used as input to predict the
ure concentrations for the non-California urban ozone
-attainment areas. The model is primarily a nationwide
el. City-specific information was only used as input for
base year ozone concentration and for the ratios of NMOC to
:. Meteorological conditions for the EKMA model are based on
a from one of three cities: a) Los Angeles - for modeling
ifornia coastal cities (therefore, not used in this
.lysis), b) Denver - for modeling cities in Arizona,
or ado, Nevada, New Mexico, and Utah, and c) St. Louis - for
.eling all other areas.
2. Summary and Analysis of Comments
Both MVMA and GM commented on what they believe is an
?rly simplistic methodology used in EKMA. MVMA stated that
?ir analysis showed that even if the ozone design level for
3 areas varied little, the per capita data, as well as the
;al NMHC inventory emissions, could be very different. It
;o stated that EPA's EKMA calculations may show too large of
change in ozone design levels even for relatively small
iuctions in NMHC emissions (as with the proposed volatility
itrol progreim). Since EKMA is actually a nationwide model,
feels that the air quality projections for individual cities
:e oversimplified. It did, however, request that EPA present
? results of each city separately. It commented that by
ing such a simplistic modeling approach, EPA has
necessarily introduced considerable uncertainty into the air
ality predictions. MVMA felt that by using nationwide
arages in the model, EKMA projections should only be used in
relative, not an absolute sense. It stated that the EKMA
del was developed in order to quantify the current
derstanding of ozone chemistry. No model on ozone formation
-------
3-20
has yet been universally accepted. Therefore, it argued, the
EKMA model should only be used as guide to the development of
control strategies in a relative and not an absolute way.
The fact that the design values for two areas coulc be
similar while NMHC emissions are very different is lot
surprising. Ozone is a complex pollutant, and is dependen on
many other variables such as NOx emissions. In regard to the
comments challenging the accuracy of EKMA, it is noted :hat
this approach has been defended at length elsewhere and these
arguments will not be repeated here. Concerning the
appropriate role of the model, it is agreed that because of the
simplified nature of this approach, it should not be used in an
absolute, or city-specific, sense. Rather, its role is to
translate VOC reductions into ozone impacts using the limited
available data. As was noted, no model has been universally
accepted yet. However, it has been recognized that EKMA is a
reasonably accurate method of relating VOC emissions to ozone
impacts. Thus both the use of EKMA and the role it plays in
EPA analyses are appropriate at this time.
-------
3-21
References (Chapter 3)
1. "The Motor Fuel Consumption Model Thirteenth
Periodical Report," Prepared for U.S. DOE by Energy and
Environmental Analysis, Inc., DOE/OR/21400-H5, January, 1988
2. "Effect of Running Losses and Excess Evaporative
Emissions on Future VOC Emission Inventories," Memorandum from
Richard A. Rykowski, Senior Project Manager,
EPA/OAR/OMS/MVEL/ECTD/SDSB, to The Record, December 22, 1988.
3. Dodge, M.C., "Combined Effects of Organic Reactivity
and NMHC/NOx Ratio on Photochemical Oxidant Formation - A
Modeling Study," Atmospheric Environment, August 1984.
4. Ito, K. , et. al. , "Photochemical Reaction of
Alcohol-Fueled Engine Exhaust Gases," 7th International
Symposium on Alcohol Fuels, 1986, 433-438.
5. Singh, H.B., et. al., "Reactivity/Volatility
Classification of Selected Organic Chemicals: Existing Data,"
EPA-600/3-84-082, 1984.
6. Caplan, J.D., "Smog Chemistry Points the Way to
Rational Vehicle Emission Control," SAE Transactions, Vol. 74,
1966.
7. "The Effect of Vehicle Running Losses on Future
Ozone Non-Attainment," EPA Memorandum from Don Clay, Acting
Assistant Administrator for the Office of Air and Radiation, to
the Administrator, EPA, October 6, 1988.
-------
Chapter 4
Economic Impact
The societal cost of RVP control is composed of several
elements, all of which will be presented in this chapter.
:?irst, refineries incur a cost when reducing the RVP of the
fuel. Second,.increases in fuel energy density and reductions
in fuel evaporative emissions result in savings to the
consumer. Changes in drivability and safety also have an
economic impact on society. Finally, enforcement of RVP
regulations also results in a minor cost to society. These
economic considerations are the subject of this section.
Finally, the effect of RVP controls on crude oil imports will
also be evaluated.
I. Refining Costs
A. Synopsis of NPRM
Chapter 5 of the Draft RIA accompanying the NPRM evaluated
the economic costs and credits associated with the proposed RVP
regulations. Modeling work performed by Bonner and Moore which
evaluated the cost of RVP controls at 1 and 2 psi increments
was used in determine refining cost. Fuel survey data was used
to determine a national average baseline RVP. By applying
Bonner and Moore's modeling results, average nationwide costs
of control were determined for a number of RVP control
scenarios.
B. Summary and Analysis of Comments
The majority of comments which were recieved on the
economic impact analysis presented in the Draft RIA pertained
to the proposed 1992 regulations of 7.0, 7.8, and 9.0 psi RVP
fuel in Class A, B, and C areas, respectively; particularly the
cost and feasibility of RVP control below 9 psi. However, a
significant portion of the comments received dealt with issues
applicable to both the interim and 1992 regulations, and some
with the interim regulations specifically. Those comments will
be dealt with here.
1. Feasibility of and Leadtime for Interim Control
/ Program
There was widespread agreement among commenters, including
all refiners, that the proposed interim control program would
not present difficult technological problems for the production
of Class C fuel. Although some concern was expressed about
reductions below 9 RVP in Class A areas, no commenter expressed
concern about specific technological barriers to the proposed
-------
4-2
program. In evaluating an interim program based on Class A, B,
and C RVP standards of 9.0, 9.5, and 10.5, respectively, API
raised no concerns relating to feasibility or leadtime and
stated that this level of control would require little or no
capital investment. Most refiners either agreed with API or
were silent .on the issue of an interim program. Chevron
specifically projected that the required RVP reduction would be
made primarily by reducing the amount of butane in the blends,
and that no capital investment would be necessary.
Conversely, Texaco commented that certain refiners would
be required to spend capital investment funds for distillation
and butane disposal to comply with an interim 10.5 psi standard
in Class C, but provided no supporting evidence, and did not
identify which refiners the comment applied to. Given the
current degree of discretionary butane blending by the refinery
industry it is unlikely that this would be the case. Another
commenter, Sinclair Oil Corporation, stated that EPA's proposed
Phase I regulations would require investments of $200,000 at
two Wyoming refineries. However, as described in Chapter 2,
many of the Rocky Mountain and Central Plains states which the
Sinclair refineries service have been reclassified to less
stringent levels (the most stringent standard has been, relaxed
from 8.2 to 9.5 RVP). Thus, for many of these states, meeting
control levels will actually require no reduction in RVP from
baseline levels.
EPA continues to believe that reductions in RVP to 9, 9.5
and 10.5 psi in Class A, B, and C areas, respectively, are
achievable for all refiners without capital investment and with
very little leadtime. Such interim standards would achieve
some RVP reduction in Class B and C areas and would cap RVP in
Class A areas at current levels. A recent letter to EPA from
API stated that 45 days would be required to begin production
of controlled fuel, allowing time for changes in refining
processes and changes in arrangements regarding crude oil and
butane. The letter lacked specifics as to why these tasks
should generally take six weeks, and what refiners might do to
hasten these events. While six weeks may allow all refiners to
proceed in a business-as-usual fashion, EPA expects that some
refiners may have to modify their standard practices relating
to purchasing and distribution.
EPA believes that providing 70 days after promulgation of
the standards will allow all refiners to provide terminals with
complying fuel. Some refiners may elect to use flexibilities
available to them to minimize the time they require to make
purchasing changes and to get production going, and others may
take advantage of options in the distribution system to reduce
the refinery-to-terminal transportation time. Most refiners,
we believe, will not require unusual efforts or expense in
either production or distribution. Finally, as discussed in
Chapter 1, an additional 30 days after compliance should allow
all end users to come into compliance as well.
-------
4-3
2. Volume of Fuel Controlled
Comments were recieved on EPA's draft analysis stating
that the gasoline demand projected using the MOBILE3 fuel
consumption model for future years was too low. EPA agrees and
has based this analysis of refining costs on DOE fuel
consumption projections.[1]
3. Cost of Interim RVP Control
Several commenters provided refining cost estimates for
interim volatility controls which differed with the analyses
performed by Bonner and Moore and EPA in several areas. One
such commenter, API, included in their comments an estimate of
compliance costs for the same interim controls evaluated in
this RIA. API estimated the cost of controlling fuel RVP to 9,
9.5, and 10.5 in Class A, B, and C areas to be $540 million
annually, significantly higher than the costs presented in
section III of this chapter. Since the API analysis included
all of the major comments received on interim RVP control,
irather than deal with each commenter individually, the major
areas of difference between API's and EPA's analysis will be
presented here.
One of the major areas of difference between API's costs
and those of EPA lies in assumptions regarding the length of
the transition period neccessary for compliance with summer RVP
regulations. In their comments, API stated that low-RVP fuel
would have to be produced as early as mid-March, and further,
that fuel produced in early spring would require a
substantially greater RVP reduction, since March, April, and
May ASTM limits are significantly higher than summer levels.
API increased their refining costs by approximately $150
million to reflect this. = As discussed in Chapter 2, EPA
believes that API has overestimated the length of the required
transition period, and further that the current lag between
sales and production already necessitates that volatility
specifications for a fuel sold in one month be met by fuel
produced during the preceding month. Consequently, this
increase in refining costs is unwarranted. By eliminating this
adjustment, API's cost estimate would be reduced to $407
riillion.
API also increased their estimated costs by $109 million
to correct for improper amortization of capital equipment by
Honner and Moore, and to reflect increased operating costs due
to the need for an additional month for transition in the
spring. EPA believes that Bonner and Moore has properly
amortized capital equipment costs (which apply to long-term
control costs and not to short-term costs) and that an
additional month of transition time is unwarranted. When this
cost adjustment is excluded, API's estimated costs total $298
rail lion per year.
API also commented that an R-value of 0.6 should be used
in assessing fuel economy improvements. Correspondingly, API
increased their refining cost estimates by $43 million per
-------
4-4
year. As will be discussed in section III below, EPA's
reanalysis of this issue concludes that an R-value of 0.85
should be used and that API has overestimated refining costs in
this respect as well.
A further reason why the refining costs presented later in
this chapter are lower than those of API relates to the
baseline RVPs assumed in the analyses. In the NPRM, an average
baseline RVP level of 11.7 psi (Class C equivalent) was used in
determining costs as well as emission'reductions. More recent
analysis of survey data indicate that baseline RVP levels are
somewhat lower and should remain so in the near future. (These
are shown in Table 2-1.) As shown in Table 3-1, average
baseline RVP levels in Class C areas are roughly 11.3 psi.
Baseline levels in Class B areas are approximately 10 psi (11.5
psi Class C equivalent), while baseline levels in Class A areas
are approximately 8.6 psi (11.0 psi Class C equivalent). This
change in baseline RVP levels decreases calculated RVP control
costs by roughly $60 million per year. Adjusting for the
difference in R-value and for lower baseline RVP levels would
reduce API's costs to $195 million, which is very close to
EPA's estimate of $196 million for refining costs less fuel
economy savings (presented later in this chapter).
API also included approximately $100 million in their cost
estimates for compliance verification, the estimated cost of
testing every .shipment of gasoline. API did not present data
in support of this estimate, however, and EPA believes it to be
rather excessive. Thus, by subracting compliance verification
costs, one sees that API's refining cost estimates are actually
lower than EPA's. This is due to the fact that API
underestimated control costs by using Bonner and Moore's long
term control costs as a basis for their analysis. EPA beleives
that the short-term, "no-investment" costs presented by Bonner
and Moore are more representative for interim RVP controls, and
has used these costs preferentially in the analysis presented
below.
C. Refinery Cost of RVP Control
The interim control of summer gasoline volatility
currently under consideration (volatility limits of 9.0, 9.5,
and 10.5 psi in Class A, B, and C areas, respectively) will
increase the cost of gasoline production, but should not
require any short term capital investment (as discussed further
below). The cost to refiners will merely be that of replacing
butane, (a relatively cheap, high-octane, and high RVP blending
component) with other gasoline components and with additional
processing, to meet sales volume and pool octane
requirements. The cost of compliance presented in this
chapter should decrease in future years as refiners install and
-------
4-5
employ facilities to convert butane into a low RVP, high octane
olending component (e.g. MTBE, ETBE, etc.), unless the second
level of RVP control takes place first.
The refinery cost modeling work performed for EPA by
Bonner and Moore Management Science for this final analysis is
improved in several ways over the modeling done for the NPRM.
(The latest modeling results were placed in the docket shortly
after the NPRM was published.) First, it was possible to
incorporate directly into the model the estimated effect on
refineries of reducing the demand for gasoline under a range of
volatility-control scenarios. (RVP control reduces the amount
of purchased gasoline which is lost to evaporation, thus
reducing the volume of gasoline sold.) Second, the latest
modeling was able to estimate the impact of a drop in the price
of butane on refiners' raw material purchases and on the demand
for the natural gas liquids (NGL) industry's products. A final
improvement was to extend the range of the original modeling
(which evaluated RVP reductions of only one and two psi) to
evaluate reductions of one, two, and three psi. (This latter
change has improved the accuracy of EPA's estimates of refinery
costs at lower levels of RVP control.)
In addition to having Bonner and Moore improve the
modeling itself, EPA modified the modeling results in three
important ways to better reflect reality. First, we excluded
the results for Bonner and Moore's Region 4 (California)
because the model's predicted base case fuel contained
unrealistically high levels of butane and low levels of
pentane. Second, the estimate of the reduction in gasoline
volume due to recovered evaporative emissions was revised to
reflect the new higher evaporative emission factors. Third,
case results were adjusted to reflect the $20 per barrel crude
cost used in the NPRM.
EPA also applied the results of the modeling in a more
sophisticated manner. Using the revised state-by-state RVP
baseline levels and standards discussed above, EPA determined
the current and final RVP level of each state's fuel by month
for each control scenario and applied a refinery cost to each
oase. This allowed the determination of a separate cost for
each of the three control levels (9.O., 9.5, and 10.5 RVP). As
described in Section II, initial RVP levels thus calculated
were somewhat lower than those in the NPRM.
Table 4-1 shows results of the Bonner and Moore modeling
for Regions 1, 2, and 3 as described in their study. The cost
estimates shown represent the additional cost incurred by the
refinery in producing low RVP fuel (the cost to the refinery of
producing the control case volume of reduced-RVP gasoline less
the cost of producing the. base case volume of base-RVP
gasoline). This cost can theoretically be separated into three
-------
4-6
Table 4-1
Bonner and Moore RVP Control Costs
(W/ Investment, $22/bbl crude, excluding California)
RVP Reduction Level
Region 1
Pool Avg RVP (psi)
Gasoline Volume (MBPD)
Direct Refining Cost (M$/D)
Direct Refining Cost ($/bbl*)
Region 2
Pool Avg RVP (psi)
Gasoline Volume (MBPD)
Direct Refining Cost (M$/D)
Direct Refining Cost ($/bbl)
Region 3
Pool Avg RVP (psi)
Gasoline Volume (MBPD)
Direct Refining Cost(M$/D)
Direct Refining Cost ($/bbl)
National
Pool Avg RVP (psi)
Gasoline Volume (MBPD)
Direct Refining Cost (M$/D)
Direct Refining Cost ($/bbl)
Base
11.
648.
—
"
11.
2075.
—
^~^~
10.
3132.
—
"
11.
5857.
—
—
75
94
01
17
92
91
04
02
1st
10.
645.
11.
0.
10.
2064.
173.
0.
9.
3116.
370.
0.
10.
5825.
555.
0.
73
52
0
017
05
17
4
084
97
30
8
118
08
99
2
095
2nd
9.
642.
94.
0.
9.
2055.
664.
0.
9.
3103.
926.
0.
9.
5802.
1685.
0.
71
91
6
146
09
87
7
320
02
77
4
296
12
55
7
288
3rd
8
640
259
0
8
2047
1339
0
8
3091
1554
0
8
5779
3152
0
.69
.38
.1
.399
.14
.78
.6
.646
.07
.55
.0
.496
.16
.71
.7
.538
Denominator is barrels of gasoline produced in base case
-------
4-7
individual components: 1) the cost to the refiner of reducing
the RVP of the entire base volume of gasoline to the control
level, 2) the savings to the refiner in production costs
resulting from a slight fuel economy increase and corresponding
volume reduction, and 3) the savings to the refiner resulting
from evaporative emission reductions and the corresponding
decrease in gasoline volume requirements.
For the purpose of estimating RVP control costs for
different ASTM classifications, it is advantageous to determine
the first component of the cost separately, and to subsequently
take credits for fuel economy improvements and emission
reductions. To do this, the cost savings to the refinery
resulting from the reduced gasoline volume requirements was
determined by multiplying the gasoline volume reduction in each
control case by the pool-average incremental gasoline cost. By
adding this value to the costs shown in Table 4-1, the cost of
controlling RVP (excluding volume reduction savings) as a
function of RVP levels was determined.
Several adjustments were subsequently made to these
values. First, costs were adjusted to a crude oil price of $20
per barrel. Bonner and Moore had evaluated the sensitivity of
RVP control costs to crude oil price with modeling runs in
Region 3. Cases were run under $22, $17, and $27 per barrel
crude scenarios. By interpolation, the cost of RVP control at
$20 per barrel was determined for Region 3, and costs for
Regions 1 and 2 were adjusted proportionally.
Second, Bonner and Moore also ran cases for Region 3 which
determined the cost of RVP control assuming no investments were
:tiade by refineries. Since interim RVP regulations will require
compliance before significant investments can be made, these
no-investment case results are more representative of the short
term situation. Thus, the relative cost of fuel control in
Region 3 under the "no-investment" and "investment" scenarios
was evaluated. Next, a proportional adjustment was made to
"investment" control costs in Regions 1 and 2. The end result
of these adjustments were "no-investment, $20 per barrel crude"
RVP control costs defined for three different levels of RVP
control. The national average costs of RVP control are shown
i'n Figure 4-1.
Costs taken from this curve were applied on a state by
state basis to estimate the national cost of the interim
regulations. EPA estimated the current RVP of each states fuel
by month from MVMA fuel survey data. [2] As described in
Section B.3. above, these values were lower than those used in
the NPRM. The control RVP level was then determined for each
state according to state-by-state RVP standards, as revised
from the NPRM (see Chapter 1). An RVP control cost was then
applied to the volume
-------
13
53
I
3.00
2.50
-0.50
12.00
Figure 4-1
RVP Control Costs
($20/bbl crude, excluding California)
11.00
a
RVP
No Investment
OO
-------
4-9
of fuel sold in each state over the entire compliance period of
iroughly five months, using EIA estimates of monthly sales of
petroleum products.[3] Costs were aggregated according to
volatility control classification for use in determining class
specific cost-effectiveness.
Results of the analysis showed that in those areas which
would be required to meet the 9.0 psi Class A standard, no
control costs would be incurred. This is because fuel
currently sold in Class A areas in slightly less than 9.0 psi
already, and because all Class A/B areas have been reclassified
as Class B. For fuel sold in areas required to meet the 9.5
psi Class B standard, an average compliance cost of 0.57 cents
per gallon was calculated. Compliance costs vary in Class B
areas, however. In those areas formerly defined as Class A
which have been relaxed to Class B, no cost will be incurred.
At the other extreme, in B/C areas, fuel required to comply
with Class B regulations will cost an additional 1.63 cents per
gallon. Average compliance costs for those areas required to
meet the 10.5 psi Class C standard will be 0.56 cents per
gallon, with variations similar to those in Class B. The total
::efinery cost of the interim standard will be 0.54 cents per
gallon, or $247 million per year.
II. Fuel Economy and Evaporative Emission Recovery Credits
A. Synopsis of the Draft Regulatory Impact Analysis
The DRIA examined the fuel economy impact associated with
volatility control. Specifically, since the heat of combustion
of gasoline is projected to increase with volatility control,
the fuel economy of the gasoline-fueled fleet is expected to
have a proportional increase. Increasing fleet fuel economy
has a cost savings associated with it, and as a result, this
was evaluated in the DRIA.
1. Relationship of Fuel Volatility to Heat of Combustion
The relationship between fuel RVP and the fuel heat of
combustion involves complex refinery modeling, since octane and
other performance requirements must be maintained as the butane
and other light hydrocarbons are removed from the fuel, as well
as due to the wide range of commercial fuel heats of combustion
for any given fuel RVP. The results of a linear programming
model developed by Bonner and Moore were verified using MVMA
fuel survey data and used to estimate a 0.25 percent increase
in fuel heat of combustion with a 1 psi reduction in fuel RVP,
and a 0.56 percent increase with a 2 psi reduction.
-------
4-10
2. Relationship of Heat of Combustion to Fuel Economy
If a vehicle were 100 percent effective in utilizing the
additional energy available in the fuel due to volatility
control (no change in vehicle energy efficiency), the fuel
economy would be expected to increase 0.25 percent for the 1
psi reduction in RVP and 0.56 percent for the 2 psi reduction.
Due to heat losses in the engine and certain vehicle
operational characteristics, a vehicle typically is less than
100 percent effective in utilizing additional energy in the
fuel. The result is that the percent change in fuel economy is
less than the percent change in fuel heat of combustion
associated with fuel volatility control. This ratio is defined
to be "R".
In an attempt to evaluate R, a review of available data
showing both fuel economy and fuel heat of combustion from
testing with different fuels was performed. However, for a
variety of reasons, this test data could not be used to define
a representative value for R. Among these reasons were:
1) The large error in fuel economy and fuel heat of
combustion measurements (at best 1.0 percent and 0.9
percent respectively) compared to the maximum
expected change in fuel economy of 1.6 percent, and
the resulting wide . scatter in the R values
calculated;
2) The age of much of the data and the corresponding
ability of newer closed-loop fuel-injected vehicles
to better take advantage of differences in the fuel;
3) The fact that typically the data was collected for a
purpose other than to determine R, resulting in
variables atypical of just volatility control
changing during testing;
*
4) Test fuel differences inconsistent with those
expected to result from volatility control of
commercial fuel (e.g., Large differences in fuel
sulfur levels, aromatic levels, distillation curves,
etc.) causing fuel economy changes independent of
the fuel heat of combustion differences;
5) Testing of vehicles on different fuels without first
modifying the vehicle to react properly to those
fuels (i.e., testing with very high volatility fuels
at high temperatures resulting in overwhelming of
the evaporative control system, or testing with very
low volatility fuels under very low temperatures
resulting in poor drivability).
-------
4-11
Because of the difficulties in using available test data
to determine the appropriate value for R, a theoretical
c.nalysis of the combustion of fuel in the engine was
performed. The model selected analyzed heat losses in the
eingine as a function of not only changes in the fuel heat of
combustion, but also the mean effective temperature of the
cylinder charge, the fuel-air ratio, the fuel specific gravity,
the change in air volume reguired, and the specific heat of the
cylinder charge. The model was then developed further to take
into account the average efficiency of a gasoline vehicle, the
effects of coast and idle operation, the effects of changes in
engine friction, and the effects of changes in fuel pump
losses. Other things such as cold starting on a lower RVP
fuel, test cycle transients, open loop warm-up operation, and
possible changes in fuel/air mixture homogeneity and
distribution were found to either have no affect on the value
of R, or to affect it in a small but unquantifiable way.
The result of this analysis was an R value calculated to
be approximately 0.82 using the fuel changes expected with
volatility control. This is well within the range of the
civailable test data mentioned above. In addition, the
theoretical R value using the model and the average value from
test data on propane matched extremely well. Information on
methanol vehicle efficiencies was then used to determine a
value for R of 0.95. Due to the large differences in fuel
economy and heat of combustion between methanol and gasoline,
the errors in measurement become less significant, and the
result more reliable. As a result, this value was selected as
an upper bound for R, while 0.82 was kept as a lower bound.
3. Relationship of Gasoline Volatility to Fuel Economy
When the relationship between the fuel volatility and
energy density of the fuel was combined with our best estimate
f:'or the relationship between the energy density and fuel
6>conomy, the relationship of gasoline volatility to fuel
economy was determined. The result • was that the 0.25 percent
increase in fuel energy density for a 1 psi reduction in RVP
was estimated to result in a 0.205 to 0.238 percent increase in
fuel economy, while the 0.56 percent increase in energy density
for a 2 psi reduction in RVP was estimated to result in a 0.459
to 0.532 percent increase. When these fuel economy increases
were applied nationwide, the fuel saved was credited at a
retail price (minus taxes) of $0.82 per gallon of gasoline.
-------
4-12
4. Relationship of Evaporative Purge to Fuel Economy
In addition to the changes in the fuel causing a small
change in the fuel economy of the vehicle, the efficiency of
the vehicle at burning the vapors (assumed to be butane) purged
from the evaporative control system was also determined.
Assuming that the vehicle can compensate for the purge to
maintain stoichiometry, and applying the range in the R value
discussed above, the vehicle was found to be anywhere from
0.976 to 3.492 percent more efficient at burning the butane
from the purge, than from burning the gasoline directly from
the tank. These percentages were then applied to estimates of
the reductions in evaporative HC emissions with RVP control as
described in Chapter 5 of the DRIA.
B. Summary and Analysis of Comments
1. Relationship of Fuel Volatility to Heat of Combustion
No comments were received on this aspect of the analysis.
However, based on more recent work with the Bonner and Moore
model, new estimates for nationwide increases in heat of
combustion with volatility control have been made. Following
the refinery modeling described in .Section 4-1, volatility
control across the country to 10.5 psi in Class C areas (9.5 in
Class B areas, and 9.0 in Class A areas) results in a 0.160
percent increase in the heat of combustion of gasoline (0.0,
0.108, and 0.199 for Class A, B, and C areas respectively).
2. Relationship of Heat of Combustion to Fuel Economy
There were a number of comments on the proper use of the
available test data to determine the appropriate R value for
volatility control. In the DRIA the available test data were
determined to be unacceptable for defining an appropriate value
for R, and instead were used merely to show that the value
derived through theoretical modeling was in the range of
possible values. Certain comments from MVMA, GM, and API tend
to support our reasons for this decision, although none of them
agree with eliminating the use of the test data entirely in
f.avor of theoretical modeling.
GM stated that combining the data from the different
sources ignores the effects on fuel economy of changes in fuel
properties other than the heat of combustion. They conclude
that each data set (since it is real world) should be analyzed
individually to determine if there were too many variables
changing, not enough data available, or if the technology
represented was too old to accurately define the current value
of R.
-------
4-13
We agree with GM that real world data is always preferable
to theoretical modeling when reliable data exists.
Unfortunately, in this situation reliable data did not exist.
All of the available data, aside from being severely inaccurate
due to the error associated with fuel economy and fuel heat of
combustion measurement, also tends to be unrepresentative of
the situation we are trying to model - fuel economy improvement
associated with a reduction in fuel RVP. The majority of
available data was collected using fuels drastically different
in fuel sulfur level, specific gravity, aromatics content,
sjtoichiometric fuel/air ratio, or boiling range, compared to
that which would be expected with volatility control. In
addition, testing was often done under conditions atypical of
average in-use vehicle operation, such as using low RVP fuels
at low temperatures and high RVP fuels at high temperatures,
causing drivability problems which could lower any measured
value of R, or using vehicles which are unrepresentative of
today's electronically controlled fuel-injected vehicles. The
result is that for the available data sets, there are typically
too many variables changing, not enough data collected, and/or
obsolescent technology represented to accurately define a value
for R representative of volatility control.
MVMA pointed out a number of problems with a number of the
available data sets, showing that these studies included far
more variables than just the heating value of the fuel. As a
iresult, they concluded that aggregating the data from all
sources by model year was devoid of any statistical or
technical rationale, and went on to say that none of the data
would support a change from the R value of 0.6.
EPA agrees with this comment by MVMA in general, however,
the value of 0.6 suffers from the same limitations they
mention. The test data which determined this value was from
1970 and 1972 vehicles which could not be maintained at
stoichiometry (i.e., freguently ran rich which tends to
decrease R) . The fuels used in the testing were not production
gasolines, and as a result had significant differences in
percent aromatics, specific gravity, and distillation curves in
addition to the heat of combustion differences. In addition,
the heat of combustion was not measured, but estimated based on
other fuel properties, and the fuel economy was measured by
weighing the fuel and measuring the distance traveled. As a
result there was a significant degree of error associated with
this testing as well even though replicate testing was
performed. The report itself states that "[B]ecause of the
limited number of cars involved, these tests can only give an
indication of 1970 and 1972 car performance in general."
Calculation of R from the small changes in fuel economy and
heat of combustion is hardly a "general" observation. The
bottom line is that MVMA is correct in stating that the
available data cannot support a change from the R value of 0.6
-------
4-14
which they support. However, neither can the available data be
used to support the R value of 0.6 (or any other value). As a
result, EPA did not use the data either individually or
aggregated to determine R, but instead selected theoretical
modeling to determine the R value associated with volatility
control.
In their effort to support the. use of test data as opposed
to theoretical modeling, API stated that experimental data
properly interpreted and sufficiently representative of actual
conditions provides the best estimate of R, since theoretical
mechanisms are not well understood. They went on to say that
reliable experimental results can be generated when test
variability exceeds the measured effect by multiple testing of
certain combinations of parameters, and that it is possible to
determine cause and effect relationships from limited data
sets. They then stated that although the data on recent model
vehicles suggests that the R value may be higher for newer
vehicles than that measured on older vehicles, the data is too
limited to place any confidence in this assertion. Instead,
they emphasize that Chevron data on 1973-6 model year vehicles
is the most acceptable since it was done using commercial
fuels, and that a value of 0.6 for R based on this data is the
most appropriate estimate.
We also agree with API that test data is generally the
most reliable source of information. However, we do not
believe, nor did they establish, that any of the available data
met the stated conditions of being sufficiently representative
and of having included multiple testing of certain combinations
of parameters. As stated above, none of the data sets meet
these requirements. This includes the Chevron data, which
suffers from unrepresentative levels of aromatics and sulfur in
the fuels, differences in distillation curves and API gravity,
vehicles unrepresentative of today's vehicles, and vehicle
testing aimed at poor drivability which may affect R.
Theoretical analysis even with all of its uncertainties appears
to be the best available option.
In addition to comments on the use of test data, a number
of comments were also received concerning the accuracy of the
theoretical model developed to evaluate R. Although there were
a few comments concerning possible errors in the modeling, the
majority of these comments focused on parameters the commenters
believed would affect the value of R which were not taken into
account in the model.
API reiterated comments on the volatility study which
questioned the effects of cold starting, idling and braking,
and transient operation on the value of R. All of these have
already been discussed at length in the DRIA. Cold start was
determined to cause a likely decrease in the value of R, but
-------
4-15
was not found to be quantifiable. The effects of idling and
braking have a similar directive effect, and were, if anything,
overcompensated for in the model, and transients were
determined to have 'little or no effect on the value of R.
Thus, there appears to be a good balance in the unquantified
nature of these parameters.
In addition, API made the comment that the poorer
driveability associated with lower volatility fuels will result
in lower fuel economy and therefore, lower R values. For this
statement to be true, not only must the level of volatility
control being proposed result in overall poorer vehicle
drivability, but the resulting difference in driveability must
also significantly affect the value of R. In answer to the
first part, it is not necessarily true that a fuel volatility
reduction of the magnitude proposed will result in overall
poorer driveability. Few vehicles (mainly older, carbureted,
and improperly maintained vehicles) are expected to experience
noticably poorer cold-temperature driveability on fuels
controlled to the levels proposed by EPA. In addition, any
possible detriment at cold temperatures will likely be negated
by improved driveability on vehicles (mainly older, carbureted
vehicles) at high temperatures. The result is that even if
driveability affects the value of R, the direction of the
offect is uncertain, and is likely negligible. This especially
true for the level of volatility control pr'oposed in this
interim rulemaking. As discussed in Section 4-III, no
detriment in vehicle cold-temperature driveability is expected
with this level of volatility control, only an unquantified
hot-temperature driveability improvement.
In answer to the second part of the comment, although poor
driveability probably tends to decrease the value of R, for
driveability to affect it significantly the reduction in
driveability will have to be much more pronounced than that
expected to result from volatility control. Data provided by
API in their comments on the DRIA tends to support this
statement. Testing on 6 recent model vehicles at 3
temperatures ranging from 42°F to 80°F, and on 4 fuels ranging
in RVP from 6.5 to 10.5 resulted in no correlation between the
level of drivability demerits and the R value due to volatility
control (there was a more pronounced effect of temperature).
As a result, it is doubtful that driveability effects resulting
from the range of volatility control in question will have any
substantial impact on the value of R.
API also discovered a small error in the model. For the
calculation of the gas flow per unit time which was an input
into the model, the thermal efficiency of the vehicle was
assumed to remain constant for all cases at 38 percent.
However, in order for R to be anything other than one, the
thermal efficiency must change. Fortunately, the effect of
-------
4-16
assuming thermal efficiency constant for this purpose over the
range of change in the actual thermal efficiency expected is
negligible, and the model's outputs remain the same. Since
there is no significant effect, and since changing the model to
correct this error would make it even more complicated, no
change will be made.
GM made the comment that the hydrogen to carbon (H/C)
ratio and heat of combustion of the fuel (in BTU/lb) should
also be taken into account in the modeling, since the H/C ratio
affects the heat losses from the engine. Both of these points
are true to some extent, and that is why the heat of combustion
per pound of fuel has already been taken into account in the
model. The H/C ratio, on the other hand, is only a secondary
measure of other primary changes which take place with the
fuel. As a result, the influence of H/C on the heat of
combustion of the fuel has already been taken into account
directly by dealing with the heat of combustion itself. As
stated by GM the H/C ratio also affects the stoichiometry of
the fuel. The H/C ratio, however, proves to be a rather poor
measure for moles of products per mole of reactants (which is
how GM tried to use it in their comment). Although the H/C
ratio decreases with volatility control, stoichiometric
analysis using the fuel properties shown in Table 4-2 reveals
that the ratio of moles of products to moles of reactants
actually increases slightly with volatility control. As a
result, any additional changes to the modeling based on GM's
comment and the above discussion would if anything increase the
value of R. The magnitude of the difference is not significant
enough, however, to warrant a change to the model.
API raised a related issue which they felt should also be
taken into account in the calculation of R. The maximum
pressure rise and resulting work output from the engine is a
direct function of the number of moles of working fluid (in
addition to temperature). However, for equivalent speed and
power, a higher energy density fuel requires less fuel which in
turn results in fewer moles of working fluid. Therefore, this
will tend to decrease the value of R. In addition, the
reduction in fuel required results in a lower demand for air.
This in turn results in higher throttling losses which also
will decrease the value of R.
Stoichiometric evaluation of both 11.5 and 9 RVP fuels
using the calculated fuel properties shown in Table 4-2,
reveals API's comment to be true, at least in part. If we
assume constant energy input into the cylinder, the 9 RVP fuel
requires approximately 2.94 percent fewer moles (0.73 percent
fewer gallons) of fuel resulting in a slight (0.05 percent)
decrease in the number of moles of combustion reactants and
products. However, there is actually a very slight increase in
the amount of air required for stoichiometric combustion.
-------
4-17
Table 4-2
Fuel Properties
Butane 9 RVP
Fuel Parameter
Heat of Combustion:
BTU/lb 19643
BTU/Gallon 95661.4
I3TU/mole 1141729.7
Avg Molecular Weight 58.124
Avg Chemical Formula C4H10
H/C Ratio 2.25
Molar A°/F° Ratio (Stoic) 30.94
Mass base F/A (Stoic) 0.065115
Specific Gravity 0.5836
Density (Ib/Gallon) 4.87
Charge Specific Heat
(BTU//lbm °R) 0.2847
18500
114330
2053618.4
111.006
C8H14.8
1.85
55.692
0.069087
0.7405
6.18
0.2826
11.5 RVP*
18540.9
113489.9
1993233.3
107.505
C7.735H14.482
1.87
54.053
0.068937
0.7335
6.121
0.28267
* Values calculated based on the replacement of 1.8
volume percent of liquid gasoline with butane per 1
psi increase in fuel RVP. -
Note: Some values may vary slightly from those in the
DRIA. Those shown here are thought to be more
internally consistent since they are calculated
based onstoichiometry.
-------
4-18
This effect can be analyzed by using a simplified model.
If we assume that miles driven are proportional to the moles of
products of combustion, then the fuel economy increases by
appoximately 0.685 percent (over an assumed baseline of 28.35
mpg with the 11.5 RVP fuel) for a 0.74 percent increase in the
heat of combustion of the fuel per gallon. This yields an R
value of 0.93. Therefore, API was correct in their comment
that the R value should decrease due to a decreasing number of
moles of working fluid (assuming constant energy input).
However, since this is a different modeling technique than that
done by EPA in the DRIA, it is a decrease below l.o, not a
decrease below the EPA values. (EPA's model looked at a change
in the energy required to go one mile. This model looks at a
change in distance traveled assuming a constant energy input.)
This value is then reduced by other effects such as idle
operation just as the initial R value of 0.931 from EPA's model
in the RIA was reduced. The result that the R value from this
new model is very similar to that calculated in EPA's earlier
modeling. As a result, no change will be made to EPA's
existing model due to this comment.
Contrary to API's comment, the effect of throttling losses
does not decrease the value of R. As stated above, although
less fuel is required at lower RVP, slightly more air is
required, tending to decrease throttling losses and increase
the value of R. However, the small change in air required, and
the small percentage of throttling losses to the total energy
consumption of the vehicle (only a portion of the 6 percent
allocated for pumping losses in the DRIA) results in no
significant changes to the value of R.
In another comment, API questioned the appropriateness of
using experimental values from methanol and propane to verify
the model. The intent in using these fuels was to provide
significant differences in the measured heat of combustion and
the measured fuel economy. As a result, the errors in fuel
economy and heat of combustion measurement were small in
comparison to the differences in those properties between the
fuels. Thus greater confidence could be placed in the R values
calculated using these fuels than when two gasolines were used
which only had slight differences in properties.
Unfortunately, while a greater level of confidence can be
placed in the accuracy of the R values calculated if fuels as
widely different from gasoline as propane and methanol are
used, the R values calculated are no longer representative of
those expected from volatility control. Many parameters other
than the fuel heat of combustion affect fuel economy, and many
of these are taken into account in the model used to determine
R. However, the differences in these parameters resulting from
volatility control are very different from those which exist
between gasoline and propane or gasoline and methanol.
Therefore the only accurate means of determining an R value
-------
4-19
Associated with volatility control is to use only those changes
in fuel properties and vehicle operating parameters expected to
result from volatility control. As a result, EPA will not use
•:he results from comparing propane or methanol to gasoline as a
means to determine a representative value for R for the case of
gasoline volatility control.
In the process of evaluating these comments, it was
discovered that the portion of the model dealing with engine
friction was in error. Since engine friction is expected to
decrease due to an increase in the power output per stroke, an
increase in the energy density of the fuel should increase the
value of R. The effect of the previous modeling was just the
opposite. Upon reanalysis, it appears that the power output
per stroke is proportional to the heat energy per mole of
reactants in the cylinder, as well as to the heat losses from
the cylinder walls, and not proportional to the heat energy per
pound of air in the cylinder. The effect of this change will
be shown in the conclusion of this section.
3. Relationship of Gasoline Volatility to Fuel Economy
There were no comments on this portion of the analysis,
and as a result, the only changes made are those which result
from the changes in the previous sections.
4. Relationship of Evaporative Purge to Fuel Economy
In the DRIA EPA took a credit for the energy value of
evaporative emission reductions which resulted from its
volatility control program. To do this EPA multiplied the
emission reductions by a recovery factor or combustion
efficiency of the vapor in the engine. Ford raised the comment
that the manner in which this evaporative recovery factor was
applied was in error. They stated that the evaporative
recovery factor as determined is only applicable to the
evaporative emission reductions which result from vehicle
control (i.e., are captured by the canister and burned in the
engine). The evaporative emission reductions which result from
fuel control (i.e., uncontrolled vehicle evaporative emissions,
running losses, evaporation from storage and handling, etc.)
directly replace fuel going into the vehicle, and as a result
should be credited on a one for one basis with liguid gasoline.
Ford is correct; the manner in which the evaporative
recovery factor was applied in the DRIA was in error. Only
those emission reductions which result from vehicle control
should be multiplied by the evaporative recovery factor since
they are the only emissions which are burned in the engine as
purged vapor. Emission reductions which result from fuel
control should not be multiplied by this factor, since they are
not burned in the engine as purged vapor. As stated by Ford,
-------
4-20
the appropriate "factor" to be applied to the emission
reductions resulting from fuel control is 1.0, since they are
converted to liquid fuel. The appropriate factor to be applied
to the emission reductions resulting from vehicle control,
however, received a number of comments.
MVMA stated in their comments that instantaneous control
of the air to fuel ratio is not possible due to the broad range
of control needed for the range of vehicle operation. As a
result, the ratio of the combustion efficiency of the purged
vapor compared to fuel from the tank should be less than 100
percent. Ford echoed this statement, but gave different
justification. According to Ford, vehicles often purge the
canisters during periods when the vehicle is not doing useful
work (i.e., idling). As a result, they estimate the combustion
efficiency of the purged vapor to be only 50 to 60 percent of
that for whole gasoline.
EPA staff do not believe Ford's comment to be relevant,
since very few vehicles purge during idle, and even for the
small fraction that do, the fuel consumed during idle is only a
small fraction of the total fuel consumption of the vehicle.
The concern expressed by MVMA has some validity. Today's
vehicles typically use exhaust oxygen sensors to maintain
engine operation at stoichiometry. When a slug of hydrocarbon
from a canister enters the engine, there is a time lag between
when the engine burns the hydrocarbon, and when the oxygen
sensor reads a low oxygen level. As a result, until the
vehicle can correct back to stoichiometry the vehicle will
operate rich, and not achieve optimum use of the vapors from
the canister. Improved purge solenoids which slowly increase
the purge rate to the engine instead of allowing a "slug" of
hydrocarbon to enter the engine are expected to go a long way
in improving the engine's effectiveness at burning the purged
vapors, but nevertheless the vehicle will still operate
slightly rich during the time lags associated with the number
of small increases in purge. (For this rulemaking, EPA has
assumed that these more sophisticated and slightly more costly
purge solenoids will be used on future vehicles to meet the
exhaust emission standards.) Although this can be used to
explain why the actual combustion efficiency of purged vapor
compared to liquid fuel may be less than that which is
theoretically possible (determined to be 1.00976 to 1.03492 by
EPA in the DRIA), it does not help quantify the effect.
GM also supported a reduction in the assumed combustion
efficiency of purged vapor. They performed 22 repetitive tests
on a prototype vehicle equipped with an onboard refueling
canister in which they measured simultaneously over the 1972
FTP the fuel economy and the mass of hydrocarbon purged to the
engine. On the basis of this testing they estimated the
-------
4-21
combustion efficiency of the purged vapor including any
improvement due to R to be 90 percent, and suggested 80 percent
a.s a value practical for real world application.
Unfortunately, GM provided little supportive information
describing how the testing was performed, and no supportive
information for why the real world value should be so much
lower. Nothing was said as to the vehicle tested, the type of
purge control system, the method used to measure the mass of
the purge, as well as many other important criteria for
evaluating the data. It is assumed that GM performed the
testing on a vehicle equipped with a current evaporative
control system, and a saturated onboard canister. The onboard
canister GM used was likely larger than that required by EPA
for the evaporative controls proposed in August of 1987.
Purging this saturated canister would result in a worst case
flow of hydrocarbon vapor to the engine. Since it is likely
that the combustion efficiency of this purged vapor decreases
once the amount and concentration of HC purged to the engine
exceeds the capability of the engine to fully adjust, the low
combustion efficiency determined by GM using a vehicle with a
"stock" purge control system and an oversize canister appears
reasonable. However, EPA expects manufacturers to introduce
improved purge control hardware and software on their new
vehicles to eliminate the majority of operation outside of the
vehicle's ability to properly adjust for the purge. EPA
oxpects this to be true even under the current proposals which
will likely require larger canisters to control running losses
and other currently uncontrolled evaporative emissions.
The hardware and software necessary to accomplish this
task appears to already exist as is evidenced by the
performance of vehicles in EPA's running loss test program.
Some of the vehicles, as shown in Table 4-3, appeared able to
effectively handle the amount of vapor being purged even with
the use of current high RVP fuels at high temperatures (well
above the 95°F starting temperature). The vehicles maintained
low emission rates without experiencing elevated exhaust HC and
CO emissions (indicative of stoichiometric operation). As a
iresult, it can be said that vehicles exist today that appear
able to handle large quantities of vapor without deviating from
stoichiometry and most likely without encountering fuel economy
detriments. Thus, it appears that the ratio of the combustion
efficiency of purged vapor to liquid fuel is likely to be close
1:o the theoretical maximum for new vehicles designed for such
conditions. In addition, these vehicles were tested under
conditions of elevated temperature which are much more severe
than the July-average temperatures used by EPA to determine the
evaporative recovery credits. Using July-average temperatures,
the evaporative control systems on the vehicles would have been
Less severely stressed, and as a result, would likely have
maintained stoichiometry even better.
-------
Table 4-3
Running Loss Test Program Results
Veh No
255
256
259
260
261
Temp °F
95
95
95
95
95
Fuel RVP
11.7
11.7
11.7
11.7
11.7
Exh HC (q/mi)
0.521
0.285
0.292
0.365
0.534
1st LA4
CO (g/mi)
5.415
5.152
3.307
4.617
6.872
Running Losses
0.36
0.00
0.06
0.04
0.04
Exh HC
0.377
0.179
0.216
0.260
0.384
2nd LA4
CO
3.988
3.781
2.962
5.377
6.547
RL
2.59
0.00
0.01
0.02
0.61
i
N)
NI
-------
4-23
In the DRIA, EPA assumed, based on R value calculations,
•chat the combustion efficiency of the purged vapor was slightly
greater than 1.0. The theory of purge control, and the
subsequent combustion in the engine, suggests that this is a
maximum, and that the real world value is probably something
less. GM's data probably represents an overly pessimistic
conditions; thus, the evaporative recovery factor is probably
greater than 0.90. As a result, a realistic option may be to
assume the combustion efficiency of the purged vapor compared
to liquid gasoline (the evaporative recovery factor) to be
L.O. Therefore, whether the emission reductions are expected
to result from fuel control or vehicle control, they will be
credited at 100 percent of the value of the liquid fuel which
they displace.
C. Conclusions
1 .
Relationship of Fuel Volatility to Heat of Combustion
The comments provided by the manufacturers resulted in few
significant changes to the R value analysis. EPA still
believes that theoretical modeling is the best available option
for estimating R due to the severe limitations on the test
data. The only change 'to the DRIA's theoretical modeling
resulted from a discovery of an error in the previous model.
The corrected model is now:
Q2 = (0.62)Q2ificyl2l +
Qcyll
Qcl
(0.13)0! * QfT * (A°/F°9 + 1) *
Qf2 * (AVF0! + 1) * Qcyll
Where:
Q =the energy required to travel one mile
Qcyl =the ratio of heat lost to the
cylinder walls to the heat of combustion
of the fuel
Qc =the heat of combustion of the fuel per
gallon
=the heat of combustion of the fuel
Qf
mole
A°/F° =the
per
1
2
molar stoichiometric air to fuel
ratio of the fuel
=the baseline fuel
=the new fuel
Solving this equation for Q2/Ql using the fuel
property data in Table 4-2 yields a value of 1.00107 which when
substituted into equations 2 and 3 from Chapter 5, Section V of
the DRIA yields an R value of 0.85, just slightly greater than
the value of 0.82 in the DRIA.
-------
4-24
Due to the unrepresentativeness of other fuels of the fuel
changes expected with RVP control/ the R value of 0.95 in the
DRIA based on methanol is probably not appropriate, and as a
result, the value of 0.85 based on the theoretical model stands
alone as the best estimate for R.
2. Evaporative Recovery Factor
The comments provided by the manufacturers resulted in
only a minor change to the evaporative recovery factor.
Comments by Ford indicated that while EPA had assumed there to
be only one evaporative recovery factor, there were actually
two: one for use with evaporative emission reductions
resulting from fuel control, defined to be 1.0; and one for use
with evaporative emission reductions resulting from vehicle
control. Following analysis of the comments, this latter
factor was reduced from a value slightly greater than 1.0 as
used in the DRIA, to a value of 1.0. Since the evaporative
recovery factors were approximated to be 1.0 regardless of the
source of the evaporative emission reduction, the treatment of
the fuel related and vehicle related emission reductions in the
cost effectiveness section of this document will be identical
(consistent with the DRIA).
3. Summary of Credits
The refinery costs presented in Section 4-1 do not include
a credit for increased fuel economy or decreased evaporative
emissions, however. Based on the R factor analysis above, and
the Bonner and Moore estimates of the fuel energy density
increases and corresponding gasoline volume demand decreases
expected with RVP control, the total reduction in gasoline
volume demand was calculated. The societal cost savings
associated with this reduction in gasoline volume will be
approximately $51 million per year. The interim regulations
will also result in evaporative emission reduction of 190,000
tons per year, derived as in the Draft RIA (616,000 tons per
year including running loss estimates, See Table 3-2). The
economic credit associated with recovery of these evaporative
emissions is a approximately $54 million ($174 million)
annually (See Table 5-1).
Ill. Drivability and Safety Impacts
A. Synopsis of Draft Regulatory Impact Analysis
1. Volatility Increases and Driveability Problems
In the period from 1974 to 1985 the volatility of unleaded
regular gasoline increased by 10 to 20 percent depending on the
area of the country. In 1985 for the first time the average
nationwide summer volatility surpassed the average ASTM
-------
4-25
recommended limits. The result has been that even though
vehicles during the same period have been designed to operate
better on higher volatility fuels, some vehicles have begun to
experience varying levels of vapor lock and fuel foaming
causing unacceptable driveability. This statement was
supported by comments on the 1985 Volatility Study, information
provided by automobile manufacturers, and also by a number of
studies in the literature which had looked into the problems of
hot temperature driveability. These studies showed that
volatility levels were indeed to a point where driveability
problems could be expected on some vehicles. At the same time,
other studies demonstrated that for the period of control and
the level of control proposed, cold temperature driveability
should not be a significant concern. This last statement was
supported by in-use information from California where
volatility has already been controlled for some time.
2. Driveability Cost Estimation
The cost associated with current driveability problems (or
the cost savings resulting from improved driveability with
volatility control) was estimated by assuming that people whose
vehicles are experiencing driveability problems were willing to
pay an extra 10 to 30 per gallon for fuel which would avoid
those problems. Temperature and population data for the 10
largest ozone non-attainment cities were used to approximate
the nation as a whole. This information was then used along
with vehicle age distributions, usage patterns and information
on the fraction of vehicles with unacceptable driveability from
Coordinated Research Council (CRC) testing to determine the
fraction of fuel sold nationwide which is burned under
conditions of unacceptable driveability. Multiplying this
amount of fuel by the 10 to 30 per gallon yielded estimates of
up to $78 million per year for the nationwide cost of poor
hot-temperature driveability.
Although EPA did not rely on it for the final cost
estimation, GM and Chrysler provided information on in-use
vehicles which also demonstrated that the costs associated with
hot temperature driveability problems were real. On the basis
of the information they provided, annual warranty costs due to
hot temperature driveability problems amount to roughly $8.1
million annually, and costs associated with vehicle design
modifications to avoid hot temperature driveability problems
have amounted to roughly $124.1 million annually. EPA decided
that not all of these design costs could be recovered if fuel
volatility were reduced, however, since they are costs that
have already been incurred to fix the problems. Still, the
estimated cost savings if volatility were reduced due to
removal of certain corrective parts from the vehicle was
estimated to be $44 million annually. Since the accuracy of
the methods used to extrapolate the information provided by
-------
4-26
Chrysler and GM into nationwide costs was unknown, these costs
were not relied on. However, they did serve to support the
costs calculated above.
B. Summary and Analysis of Comments
1. 'Hot Temperature Driveability
In general, the comments on the hot temperature
driveability analysis were somewhat limited. The oil companies
who commented on the topic of driveability were generally in
favor of EPA setting volatility standards, but the level of
control recommended was typically that of the current ASTM
levels or slightly lower. They cited reasons such as poor cold
temperature driveability, the lack of a clear benefit for
further reductions, and fuel explosivity at low temperatures as
reasons why reductions much below ASTM were not justified. The
motor vehicle manufacturers, consumer groups, and environmental
groups who commented, however, cited hot temperature
driveability problems and hot temperature/high volatility
safety problems as justification for supporting EPA's proposed
level of control, with some recommending levels even lower.
API, Texaco, and Sunoco provided the only substantive
comments on the topic of hot temperature driveability with
their criticism of EPA's use of CRC driveability studies. They
stated that the CRC hot temperature driveability test, due to
(among other things) its extremely severe nature, is not
representative of in-use operation. As a result, they stated
that comparisons of. the results from this testing are not
directly applicable to consumer acceptability of a given fuel.
They also stated that some of the testing had confounding
results, and that typically it was necessary to exceed the
recommended ASTM ratings to obtain markedly poorer
driveability. They also pointed out that Figure 5-13 of the
DRIA (from which the cost estimates of poor hot temperature
driveability were derived) was used incorrectly. They stated
that the figure, taken from a CRC report, was used as a measure
of consumer satisfaction for a given fuel RVP, while in fact
the figure was derived from CRC trained rater data.
Based on information contained in the DRIA, the comments
discussed above, and comments by other vehicle and consumer
groups, it is still apparent that hot temperature driveability
problems exist, and represent a significant cost to society.
However, EPA's earlier attempt to quantify that cost appears to
have been flawed. The vehicle manufacturer information used to
support the cost credits, however, remains as a justification
for hot temperature driveability credits with volatility
control. However, since the accuracy of these estimates on a
nationwide basis are unknown, it does not appear appropriate to
take a cost credit for improved hot temperature driveability.
Also, since the driveability cost benefit proved to have such a
small effect on the cost effectiveness of volatility control in
the DRIA, this change should have little overall effect on the
regulation.
-------
4-27
2. Cold Temperature Driveability and Safety
In comparison to hot temperature driveability, a great
deal of comments were received on the topic of potential cold
temperature driveability problems if fuel volatility is reduced
significantly below the ASTM ratings. However, since no
comments were recieved suggesting that the gasoline volatility
control proposed in this regulation will have any detrimental
effect on cold temperature driveability and safety, the
comments on this topic will not be discussed in this rulemaking.
3. Hot Temperature Fuel Safety
In addition to comments on cold temperature, low fuel
volatility safety, comments were _also received on hot
temperature, high fuel volatility safety. IIHS, CAS, and NADA
all provided comments describing the current safety hazard of
high volatility fuels during high temperature operation, and
stated that EPA's proposed volatility control will go a long
way toward eliminating this problem. CAS stated that since
L979, 71 fires, 25 injuries, and 2 deaths have been linked to
12 vehicle recalls linked to fuel system overpressurization and
fuel volatility. While . they acknowledge that vehicle design
also plays a significant role in fuel system overpressurization
and fuel spurting issues, they also "maintain that control of
in-use fuel volatility will reduce the problem.
The issue of hot-temperature safety is also an issue of
merit. Reducing the volatility of in-use gasoline should serve
to reduce the problems of fuel spurting and fuel system
overpressurization. However, it is not possible on the basis
of existing data to determine to what extent historical
hot-temperature safety problems would have been avoided or
substantially reduced in severity if fuel volatility had been
lower, much less to project this into the future. As a result,
no evaluation of possible benefits due to RVP control is
attempted here.
C. Driveability and Safety Economic Impacts
Although economic impacts of hot temperature driveability
and safety were discovered, no good method of quantifying their
effect could be found. As a result, no credit is taken for
driveability or safety impacts in this rulemaking. In
addition, since the level of volatility control in this
rulemaking is low, no cold temperature impacts are expected.
IV. Enforcement Costs
The costs of several different enforcement options were
presented in the Draft RIA. Costs for the enforcement options
evaluated were estimated to range from $0.3 to 2.3 million per
year. EPA does not expect that actual enforcement costs will
-------
4-28
differ significantly from these levels. Since these
enforcement costs are small relative to the other elements of
the RVP control (such as refinery costs, etc.), the cost of
enforcement was not included in the cost effectiveness
calculations of Chapter 5.
-------
4-29
V. Effect of Volatility Regulations on Imports
One issue which has received extensive comments is the
effect volatility controls will have on imports of crude oil.
Since RVP regulations will result in the displacement of
domestically produced discretionary butane from the gasoline
pool, purchases of crude oil will necessarily increase in
order to meet a fixed gasoline energy demand. This is
especially true in the short term, before investments can be
made to install equipment neccessary to convert butane to
MTBE, ETBE, alkylate, etc. As additional processing equipment
is installed, and as a price decrease establishes butane as a
competitive petrochemical feedstock, purchases of additional
crude will decrease, by approximately 50 percent as indicated
by Bonner and Moore modeling in Region 3.
In order to assess this issue, the maximum quantity of
n-butane rejected from the gasoline pool in order to achieve
the given RVP reduction was estimated for each of Bonner and
Moore's control scenarios, using a blending value of 65 psi
for n-butane. The energy content of this rejected butane,
less the energy content of recovered evaporative emissions,
was then compared against incremental crude purchases made in
each control scenario. Results of this analysis showed that,
on average, the energy of the incremental crude purchased
exceeded that lost by the displacement of butane by a factor
of approximately 1.8.
Given this information, an analysis of the effect of the
interim volatility standards on imports was made. For all
fuel undergoing a volatility reduction, an estimate of the
maximum quantity of discretionary n-butane which would have to
be rejected was made. A total of 12.4 million barrels of
butane per year was calculated (assuming a 5 month refining
period). However, under the interim standards, 190,000 tons
per year of evaporative emissions will be recovered, the
energy equivalent of 1.9 million barrels of butane. Thus, the
energy equivalent of 10.5 million barrels of butane,
(increased by a factor of 1.8), will be required in
incremental crude oil. This totals approximately 12.3 million
barrels per year (81,000 barrels per day) or $246 million per
year at a $20 per barrel crude price.
It should be noted that this represents a short-term
estimate of incremental crude purchases. Bonner and Moore's
modeling showed that, in the long term, as butane price falls
and refiners are able (by modifying processing) to use more
butane (in applications such as petrochemical feed,
alkylation, and isomerization) the import effect may be
halved. With increased purchase of MTBE and ETBE facilities,
imports would be reduced further still. Of course, inclusion
of running losses in the estimate of evaporative emissions
recovered would have a large effect on the amount of crude oil
purchased, reducing it to as low as 48,000 barrels per day in
the short term, and less than half that in the long term.
-------
4-30
References (Chapter 4)
1. "The Motor Fuel Consumption Model Thirteenth
Periodical Report," Prepared for U.S. DOE by Energy and
Environmental Analysis, Inc., DOE/OR/21400-H5, January, 1988.
2. "MVMA National Gasoline Survey—Summer Season",
Sampling Date - July 15,1987.
3. "Petroleum Marketing Annual, 1985, Volume 2," Energy
Information Administration, Office of Oil and Gas, U.S. DOE,
DOE/EIA-0487(85)/2, December, 1986.
0351Z
-------
Chapter 5
Cost Effectiveness
EPA's methodology for determining the cost effectiveness
of various alternative combinations of vehicle-based and
fuel-based excess evaporative emissions control programs has
evolved over a number of years. The November 1985 study of
these issues published an earlier methodology, which after
public comment and internal EPA development was revised and
used in the proposal.[1,2]
Sections I.A and I.B below describe the methodology found
in the volatility NPRM, Section II contains the summary and
analysis of the comments on that methodology, and finally,
Section III presents the cost-effectiveness of the interim
proposal of volatility control to 10.5 in ASTM Class C areas,
9.5 in B areas, and 9.0 in A areas.
I. Synopsis of NPRM Methodology
A. Basic Model
The cost-effectiveness (C/E) model developed relationships
between the costs of various alternatives and the VOC emission
reductions projected to result from those alternatives. These
C/E results were used in two ways — both to compare RVP
control with other VOC programs and to compare the various RVP
control alternatives with one another. In the NPRM the model
was used to calculate C/E for eight control- cases ranging from
11. Si RVP to 8.0 RVP in half-psi increments (i.e., summertime
control to the specified RVP in Class C areas with proportional
reductions in other areas coupled with certification fuel RVP
matched to the Class C RVP).
A key concept utilized by EPA in the volatility NPRM was
that of "incremental" C/E analysis. Because a wide range of
alternatives are available which all address excess evaporative
emissions to some degree (i.e., various levels of combined
fuel-based and vehicle-based control), it is most useful to
compare the C/E of each level of control incremental to the
immediately preceding level of control and to other control
programs. Thus the model results in a range of incremental C/E
values/which show the effect of each degree of control.
•rFor each level of control analyzed, the model calculated a
cost and an emission reduction for that increment and presented
their ratio in terms of dollars per ton of VOC reduced. The
following paragraphs discuss these emission reduction and cost
calculations.
-------
5-2
Most VOC control programs reduce emissions and incur costs
throughout the year. This is not true for the proposed
gasoline volatility control. The NPRM assumed that VOC
reductions in the summertime are most effective at reducing
ozone levels, and are hence the most valuable reductions to
achieve. In comparing various 5-month volatility control
programs to such year-round programs, it is necessary to make
an adjustment for when emission reductions occur. One way of
doing this would be to modify the cost effectiveness
calculation of other year-round programs such that only
emission reductions which occur during the ozone season are
fully credited. However, the simpler and more practical method
chosen by EPA was to modify the cost effectiveness of
volatility control to make it look like a year-round control
option. To do this, EPA used the nationwide emission
inventories presented in Chapter 3 of the DRIA which were
projections of summertime emissions over the period of a whole
year (i.e., the same as multiplying the actual 5-month summer
emissions reductions by 12/5).
EPA then made a second adjustment in accounting for
emission reductions. Because the primary purpose of VOC
control programs is to achieve reductions in ozone
non-attainment areas, EPA focused the analysis of volatility
control on reductions to be achieved in these areas. To do
this, the nationwide inventory reduction value at each control
level (from Chapter 3 of the DRIA) was multiplied by the
estimated fraction of national VMT occurring in ozone
non-attainment areas (0.395). A credit was applied for
attainment areas emission reductions, which is discussed later.
The methodology which EPA used for computing cost values
involved a summation of several terms representing various
costs and savings. The factors the model incorporated at each
level of control were the following: 1) total national
refinery costs, 2) an economic credit for control in attainment
areas, 3) a credit for improved driveability, 4) a credit for
improved fuel economy, 5) a credit for utilizing captured
evaporative emissions, 6) an economic penalty for increased
vehicle weight, plus 7) the change in vehicle cost. The
following paragraphs summarize how these factors were
determined.
The refinery-level cost came from multiplying the
per-gallon refinery cost figures from the Bonner and Moore
modeling (Chapter 5 of the DRIA) by projected nationwide fuel
consumption. Four economic credits were then calculated and
applied against this refinery cost. An assessment in Chapter 5
of the value of improved driveability as fuel volatility is
reduced led to a projected economic benefit at each level of
control which was then built into the C/E model. Next,
improved fuel economy at each control level due to fuel of
greater energy density (Chapter 5) was multiplied by national
fuel consumption and the
-------
5-3
pre-tax price of gasoline. Since the improved fuel economy
values from Table 5-19 in Chapter 5 were presented for R-values
oil both 0.82 and 0.95, both sets of values were incorporated
into the C/E model to delineate the likely range for this
economic credit.
Another aspect of EPA's evaluation which led to a credit
was the recovery and combustion of vapors currently lost to the
atmosphere. The methodology, as detailed in Chapter 5, was to
convert prevented (or recovered) VOC emissions (assumed for
simplicity to be all butane) to an equivalent amount of
gasoline on an energy basis and then multiply by two factors:
1) the pre-tax retail price of gasoline and 2) combustion
efficiency factors (Chapter 5, Section V.B.4).
The final credit was a response to the focus of the model
on non-attainment area emission reductions, as described
above. While health-related emission reductions occur where
the ozone standard is being violated, EPA assumed that there
would be some value to a program such as RVP control that could
by its nature also reduce ozone in areas currently in
attainment (e.g., by preventing crop damage, etc.). Thus, as a
way of making the C/E results more comparable to the many VOC
control programs which clearly provide little benefit outside
non-attainment areas, EPA attributed a credit to RVP control
for attainment area emission reductions. The value of
attainment-area VOC reductions had not been rigorously
evaluated by EPA, but a previously used estimate of $250 per
ton was applied to the 60.5 percent of nationwide emission
reductions assumed to occur in attainment areas.
In addition to these credits, EPA calculated and added a
small fuel economy penalty which would result from increased
vehicle weight under vehicle-based control scenarios
(Chapter 4). Finally, the cost of vehicle improvements under
the various scenarios, as calculated in Chapter 4, were added
to the other costs and credits. The sum of all these factors
a.t each level of control resulted in C/E values comparable to
other VOC control programs.
B. Types of C/E Analyses
EPA used the basic C/E model to calculate a full set of
cost-effectiveness values for two distinct analyses, a series
of single-year analyses and a 33-year average analysis.
1. Single-Year Analyses
The single-year analyses had two purposes. The first was
1:0 evaluate steady-state C/E as represented by the C/E in the
year 2010. By that time, turnover of the vehicle fleet will be
essentially complete; nearly all vehicles can be assumed to be
equipped with improved evaporative systems. This
-------
5-4
"steady-state" type of analysis thus allowed a "snapshot" of
the programs C/E once it has been fully implemented.
The second purpose of the single-year analyses was to
examine short-term RVP control options. Short-term analysis
showed for a given long-term RVP level whether temporary deeper
RVP reductions (without matching certification fuel volatility
reductions) prior to 2010 would also be cost effective.
2. 33-Year Average Analysis
The other type of cost effectiveness analysis is analogous
in most ways to the single-year analysis. The only difference
is that here C/E was evaluated for each of 33 years, including
the earlier years of the program when start-up costs would be
incurred. Costs and emission reductions were calculated for
each year to reflect, for example, the gradual turnover of the
vehicle fleet. The present value of both costs and emission
reductions was then calculated using a discount rate of ten
percent.
C. Comparative C/E Value
The NPRM used a C/E of $2000 per ton as a guideline for
comparing RVP control programs to other VOC programs. The
$2000/ton benchmark was based on a judgement of how cost
effective remaining programs likely to be required to reduce
ozone will be. EPA noted that a number of programs being
considered will have significantly higher C/E values (52 FR
31286).
D. Sensitivity of C/E Analyses to Various Factors
EPA performed several separate steady-state and 33-year
average analyses to test the sensitivity of the cost
effectiveness results to certain key assumptions. The
assumptions tested were the following: l) The presence or
absence of onboard refueling control requirements; 2) the
future cost of crude oil; 3) the use of nationwide costs and
emission reductions versus separate costs and emission
reductions for the individual ASTM class areas; and 4) the
influence of the number of non-attainment areas in which
emission reductions are projected to occur. In addition EPA
also performed the latter two analyses for the case of
$33/barrel crude oil with federal income taxes excluded.
E. Cost Effectiveness of Mid-Range Volatility
Finally, using a separate methodology, EPA analyzed the
cost effectiveness of controlling the mid-range volatility of
fuel, as measured by the percent of fuel evaporated at 160°F
(%160)- Because EPA received no specific comments about the
methodology used for determining the C/E of %ieo control, the
-------
5-5
description of that methodology found in Chapter 6, Section II
of the Draft RIA is not repeated here.
II. Summary and Analysis of Comments
This section will focus on comments directed at the
volatility NPRM's cost effectiveness methodology. A number of
issues raised by commenters affect C/E but are actually
directed at the inputs which form the foundation of the C/E
methodology. These comments are summarized and addressed in
the appropriate section elsewhere in this document (i.e.,
refinery costs in Chapter 4, emission reductions in Chapter 3,
etc.).
Similarly, several commenters re-calculated C/E results
using different input values. The results themselves are not
appropriately considered comments on C/E since they depend
entirely on the input values, which are addressed elsewhere in
this document.
A. Basic Model
While no commenters took issue with the incremental nature
of EPA's cost effectiveness analysis, there were a number of
comments on the basic model EPA used to compare the cost
effectiveness of volatility control to other control options.
1. Method of Adjusting for Full Year Comparison
One of the more significant comments involved objections
to the method EPA used to adjust the cost effectiveness values
of the 5-month volatility program to emulate those of other VOC
control programs which are year-round. Texaco, API, and others
commented that adjusting the emission reductions to reflect
year-round emission reductions without adjusting the costs,
distorts the C/E analysis and places a high value on
nonexistent emission reductions.
While EPA agrees that this adjustment distorts the actual
omission reductions available with RVP control, the C/E
-------
5-6
Therefore, modifying this aspect of EPA's C/E methodology would
itself result in a distortion of comparisons with other control
options, and no change will be made.
2. Attainment Area Credits
A topic receiving a great number of comments involved the
benefit of $250 per ton which EPA placed on emission reductions
in attainment areas. The Conservation Law Foundation and the
National Resources Defence Council supported EPA taking a
credit for emission reductions in attainment areas. Reasons
cited include: 1) benefits associated with reducing the
problems of ozone and VOC transport to and from non-attainment
areas; 2) reducing the possibility of future non-attainment of
current attainment areas due to growth in emissions; 3) the
possibility of a future lower ozone standard bringing many
current attainment areas into non-attainment; 4) possible
health benefits in attainment areas due to little or no margin
of safety associated with the current ozone standard; 5)
welfare effects such as materials, crop, and forest damage at
concentrations below the current ozone standard; 6) reduction
in the contribution of ground-level ozone to the greenhouse
effect; 7) reduction in air toxics emissions resulting from the
reduction in gasoline related VOC emissions. On the basis of
these attainment area benefits of VOC control, NRDC supported
an attainment area benefit much greater than the assumed value
of $250 per ton.
Contrary to these comments, Phillips, MVMA, API, GM, and
Chrysler all provided comments stating that the $250 per ton
credit taken by EPA was completely unjustified. The comments,
GM's in particular, focused around the lack of scientific
support for the credit, and also around EPA's lack of legal
authority to claim benefits in attainment areas.
Concerning EPA's legal authority to take a credit for
emission reductions in attainment areas, GM, MVMA, and Chrysler
appear to misunderstand either the Clean Air Act, or what we
proposed. In the NPRM, EPA clearly stated that the primary
purpose of the volatility control program was to improve
compliance with the ambient ozone standard. That is why only
non-attainment area emissions appear in the denominator of the
C/E equation. However, when EPA implements a nationwide
emission control program such as volatility control, EPA is
required to consider the costs and benefits of the program in
non-attainment areas and attainment areas alike. Therefore it
is also most' appropriate for EPA to take credit for any
benefits which occur in attainment areas, as well as
non-attainment areas. This is unrelated to EPA's legal mandate
to protect human health.
Aside from the legal authority, the question still remains
as to what value should be placed on the benefits in attainment
-------
5-7
areas. The value of $250 per ton was taken from a sensitivity
analysis performed by GM, and was not based on any scientific
data. Few if any of the commenters stated that they believed
the value to be zero, yet they presented no information to show
that it was anything other than zero. GM belittled the value
of VOC emission reductions in attainment areas, citing reasons
such as high VOC/NOx ratios, high naturally occurring VOC
emissions, non-zero background ozone • levels, as well as
transport of stratospheric ozone in attainment areas. While
most of GM's arguements are valid to one degree or another,
they do not result in a conclusion that VOC reductions in
attainment areas as a whole have no value, rather that the
value is smaller than in NA areas as is assumed by EPA. EPA
has been working on placing a value on reductions of VOCs in
ctttainment areas as well as non-attainment areas. While the
work is not yet complete, the preliminary results are
significantly higher than the $250 per ton used in the proposal
f:or this regulation. Benefits of control which EPA is
considering include:
1) Reduced ozone and ozone precursor transport to
non-attainment areas;
2) Health benefits at ozone concentrations below the
standard;
3) Welfare benefits at ozone concentrations below the
standard (materials, crop, forest damage);
4) Reduced air toxics emissions (e.g., benzene);
5) Reduced contribution to ambient particulate matter
and the problems associated with particulate
(Through condensation and formation of secondary
aerosols, VOCs can form particulate matter);
6) Odors, soiling, and morbidity associated directly
with volatilization of organic compounds emitted to
the atmosphere.
Based on these benefits a non-zero value seems appropriate, and
for purposes of this rulemaking the value of $250 will be
retained.
Also on the topic of attainment area credits, API
questioned the manner in which EPA applied the value of
$250/ton. API stated that it should be multiplied by 5/12
since the emission reductions will only occur in the
summertime. Actually, the $250/ton figure is $250 per
year-round ton of VOC reduction, or on the same basis as the
$2000/ton guideline used for NA areas. As a result, the actual
summertime attainment area emission reductions need to be
adjusted by 12/5 in order to put them on a year-round basis;
just like the non-attainment area reductions.
-------
5-8
3. General Methodology
Texaco stated that due to fuel weathering, the cost
effectiveness must be shifted to allow for a 1.5 psi margin
between marketed fuel and the fuel actually used by the
engine. The effects of weathering, however, were taken into
account when the evaporative emission factors were calculated
in the DRIA and again in Chapter 2 of this Final RIA, and need
not be corrected for again as Texaco suggests.
4. Transport Regions
NRDC stated that an alternative means of taking a credit
for some of the benefits which acrue in attainment areas might
be simply to include the transport regions and borderline
non-attainment areas into the fraction of the country currently
in non-attainment, and credit emission reductions in those
areas as in non-attainment areas. EPA recognizes the rationale
behind such a strategy, and as stated in Chapter 3, EPA is
studying the transport phenomenon. By examining the ratio of
$2000 to $250 per ton, it is clear that this approach would
only reguire one eighth of the tonnage to be eguivalent to the
$250/ton credit. This would only be 7.5 percent of nationwide
VMT, or less than 20 percent of non-attainment area VMT.
However, until the transport work is completed we believe that
the method chosen by EPA is the more justified of the two
alternatives, and will not change approaches for this
rulemaking.
B. Types of C/E Analyses
All comments provided apply more specifically to the basic
model or to one of the other areas, and as a result will not be
repeated here.
C. Comparative C/E value
A number of comments were recieved not only on the value
to be placed on emission reductions in attainment areas, but
also on the $2000 per ton used as a guideline for acceptable
cpst effectiveness of VOC control. NRDC stated that there is
rfo legal or scientific support for setting a maximum cost per
ton of control at $2000. Chrysler stated that EPA had not
shown that a cost of $2000 per ton was cost effective. Sohio
supported Chrysler's statement, and went on to state that the
$1100 per ton used by EPA in the lead phasedown regulation
should be selected as an upper bound. NESCAUM on the other
hand pointed out that many other control measures implemented
and being considered in the Northeast have costs much higher
than $2000/ton associated with them.
In so far as EPA has not performed a cost benefit analysis
which attempts to place a cost savings to society for all of
-------
5-9
the benefits which result from VOC control and ozone level
reductions, Chrysler and Sohio are correct. EPA is developing
such an analysis, but it will not be completed for some time.
Instead, EPA (as NESCAUM suggested) focused on the cost
effectiveness of other control options which might be
implemented to reduce VOC to bring non-attainment areas into
compliance.
EPA's Office of Policy Planning and Evaluation
commissioned a study which estimated the cost of current and
potential VOC control options.[3] The cost effectiveness of
these options ranged from -$3260 to $80853 per ton. If all
currently available control options were applied, many areas of
the country were found to still require additional VOC
reductions. The average cost of emission control programs to
achieve addition emission reductions was then estimated to be
within the $2000 to $10000 per ton range. The value assumed by
EPA in the NPRM of $2000/ton represents the low end of this
range, and as a result appears reasonable for use as a
guideline to demonstrate the cost effectiveness of volatility
control. However, as NRDC stated, this level is not to be
considered as an absolute ceiling for the C/E of a VOC control
program.
D. Sensitivity of C/E to Various Factors
. A few of the commenters estimated the C/E of volatility
control on their own using different values for many of the
input variables, adding to the sensitivity runs already
performed by EPA. Selection of the appropriate input variable,
however, is typically not part of the cost effectiveness model
itself, but is addressed elsewhere in the RIA. As a result
these comments will be addressed in those places. In addition
to these comments, NRDC commented that due to the uncertainties
surrounding the adequacy of the current ozone standard, EPA
should perform sensitivity runs assuming ozone standards of
0.08 and 0.10 ppm. Such sensitivity runs, although useful for
observing the effects of reductions in the ozone standard,
would not add significantly to this rulemaking since we
conclude that the program is cost effective even at the current
cuzone standard.
OMB provided comments suggesting that the cost
effectiveness of volatility control in Class A and B areas
should be examined independently of Class C areas, since the
C/E of volatility control for Class A and B fuels is greater
than Class C fuels. In Section C below, the class-specific
cost effectiveness runs are performed in addition to the
nationwide cost effectiveness.
-------
5-10
E. Alternative Programs
Comments were provided by GM and a number of other
organizations supporting various modifications or substitutions
for the regulations proposed by EPA. All of these comments
lacked sufficient analysis to justify any deviation by EPA from
its proposals. In some cases partial analyses or results of
analyses were presented. However, even with this information
there was no reason to believe that the alternatives proposed
by the commenters would result in an equal or greater level of
VOC control at an equal or lower cost.
Ill. Final Analysis
For the 1989 through 1991 timeframe, EPA will require
gasoline sold during the summer months to have its volatility
reduced below the maximum ASTM limits to 10.5 psi (primarily in
Class C areas), 9.5 (primarily in Class B areas), and 9.0
(primarily in Class A areas, no reduction). (The specific
standards for each area are outlined in Chapter 1.) The cost
effectiveness analysis for this level of control is very
similar to that performed in the proposal. However, since this
is only a short term control program, a 33-year cost
effectiveness analysis is not applicable, and no vehicle costs
are involved. The cost effectiveness analysis is performed for
1990 and' assumed to be representative of the entire 1989-1991
control period.
A. Class-Specific Emission Reduction Estimates
Class-specific emission reduction estimates were derived
in Chapter 3 and presented in Table 3-2, and will not be shown
again here. Running loss and excess evaporative emission
reduction estimates were also derived, and are included here as
a sensitivity analysis.
B. Class-Specific Control Cost Estimates
The refining cost per barrel of gasoline of controlling
the volatility of the fuel for the control case described above
was taken from Chapter 4. The refining cost for each ASTM
glass area was then determined by multiplying these per barrel
costs by the total nationwide gasoline consumption as estimated
by DOE for 1990, and the fraction of gasoline sold in each
area.[4] From these costs were then subtracted credits for:
increased fuel economy using the same methodology as in the
DRIA but with an R factor of 0.85, fuel energy density
estimates from Chapter 4, and fuel consumption estimates from
DOE; fuel recovery through a reduction in evaporative
emissions, also following the methodology in the DRIA, but now
using a fuel recovery factor of 1.0, and new summertime July
average emission reduction estimates (less exhaust emission
reductions); and an attainment area credit of $250 per ton of
-------
5-11
emission reductions occuring in attainment areas (44.6% in
Class C areas, 75% in Class B areas, and 76% in Class A areas)
using the year-round July average emission reductions found in
Table 3-2. Unlike the NPRM, no credit was taken for vehicle
driveability improvement with lower volatility fuels. In
addition, since no vehicle control will be required, there are
no vehicle costs or weight penalty costs which must be added
into the analysis. The resulting net costs of this gasoline
volatility control scenario (adjusted for summertime and
nationwide control) are shown in Table 5-1. As can be seen, if
running loss and excess evaporative emission reductions are
included, the credits are larger than the refining costs, and
the adjusted costs become negative. Since the average gasoline
currently in-use in Class A areas is less than the level of
control currently being promulgated, there are no control costs
in Class A areas.
3. Cost Effectiveness
The cost effectiveness of the control program was then
determined by dividing the net adjusted costs from Table 5-1 by
the year round design value day emission reductions occurring
in NA areas as shown in Table 3-2. The resulting cost
effectiveness of this volatility control program are well with
the $2000/ton guideline ($236/ton nationwide, $165/ton in Class
C areas, $576/ton in Class B .areas., and $0/ton in Class A
areas). If the running loss and excess evaporative emission
reductions are included, then the resulting C/E are all less
than zero (-$201/ton nationwide, -$155/ton in Class C areas,
-$425/ton in Class B areas, and $0/ton in Class A areas. Thus,
this volatility control option can be considered to be very
cost effective.
-------
5-12
Table 5-1
Cost Estimates for the Interim RVP Program ($million/yr)
Summer
Refining
Cost
Fuel
Economy
Credit
Fuel
Recovery
Credit
Net
Cost
Attainment
Area
Credit
DRIA Emission Reductions Only
Class A
Class B
Class C
Nationwide
0.0
88.56
158.03
247.09
0.0
11.70
38.95
50.69
0.0
17.11
36.42
53.53
0.0
59.75
82.67
142.88
0.0
29.64
40.17
69.81
Adjusted
Net
Cost
0.0
30.11
42.49
73.06
Including Running Loss and Excess Evaporative Emission (for Sensitivity)
Class A
Class B
Class C
Nationwide
0.0
88.56
158.03
247.09
0.0
11.66
38.81
50.52
0.0
54.72
118.78
173.51
0.0
22.17
0.44
23.06
0.0
88.39
116.81
205.20
0.0
-66.22
-116.37
-182.14
-------
5-13
References (Chapter 5)
1. "Study of Gasoline Volatility and Hydrocarbon
Emissions from Motor Vehicles," U.S. EPA, OAR/OMS/ECTD,
EPA-AA-SDSB-85-05, November 1985.
2. "Draft Regulatory Impact Analysis, Contol of
Gasoline Volatility and Evaporative Hydrocarbon Emissions from
New Motor Vehicles," U.S. EPA, OAR/OMS/ECTD, July 1987.
3. "National Assessment of VOC, CO, and NOx Controls,
Emissions, and Costs," Prepared for Office of Policy Planning
and Evaluation, U.S. EPA, by E.H. Pechan-and Associates, Inc.,
September 1988.
4. "The Motor Fuel Consumption Model Thirteenth
Periodical Report," Prepared for U.S. DOE by Energy and
Environmental Analysis, Inc., DOE/OR/21400-H5, January, 1988.
------- |